Revert "[Not to Main] Docker Release Mooncake Deps (#159)"

This reverts commit d7045619c1.

.buildkite/image_build/image_build.sh

@@ -92,8 +92,8 @@ check_and_skip_if_image_exists() {
 }
 
 ecr_login() {
-    aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin "$REGISTRY"
-    aws ecr get-login-password --region us-east-1 | docker login --username AWS --password-stdin 936637512419.dkr.ecr.us-east-1.amazonaws.com
+    aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin "$REGISTRY" || true
+    aws ecr get-login-password --region us-east-1 | docker login --username AWS --password-stdin 936637512419.dkr.ecr.us-east-1.amazonaws.com || true
 }
 
 prepare_cache_tags() {

.buildkite/image_build/image_build_cpu.sh

@@ -11,7 +11,7 @@ REPO=$2
 BUILDKITE_COMMIT=$3
 
 # authenticate with AWS ECR
-aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin "$REGISTRY"
+aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin "$REGISTRY" || true
 
 # skip build if image already exists
 if [[ -z $(docker manifest inspect "$REGISTRY"/"$REPO":"$BUILDKITE_COMMIT"-cpu) ]]; then

.buildkite/image_build/image_build_cpu_arm64.sh

@@ -11,7 +11,7 @@ REPO=$2
 BUILDKITE_COMMIT=$3
 
 # authenticate with AWS ECR
-aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin "$REGISTRY"
+aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin "$REGISTRY" || true
 
 # skip build if image already exists
 if [[ -z $(docker manifest inspect "$REGISTRY"/"$REPO":"$BUILDKITE_COMMIT"-arm64-cpu) ]]; then

.buildkite/image_build/image_build_torch_nightly.sh

@@ -46,7 +46,7 @@ echo "Image not found, proceeding with build..."
 
 # --- CUDA 13.0 for nightly builds ---
 # Nightly CI uses CUDA 13.0 while regular CI stays on CUDA 12.9
-NIGHTLY_CUDA_VERSION="13.0.0"
+NIGHTLY_CUDA_VERSION="13.0.2"
 NIGHTLY_BUILD_BASE_IMAGE="nvidia/cuda:${NIGHTLY_CUDA_VERSION}-devel-ubuntu22.04"
 NIGHTLY_FINAL_BASE_IMAGE="nvidia/cuda:${NIGHTLY_CUDA_VERSION}-base-ubuntu22.04"
 

.buildkite/intel_jobs/engine_intel.yaml

@@ -0,0 +1,21 @@
+group: Engine Intel
+depends_on:
+  - image-build-xpu
+steps:
+- label: Engine (1 GPU)
+  timeout_in_minutes: 30
+  device: intel_gpu
+  no_plugin: true
+  working_dir: "."
+  env:
+    REGISTRY: "public.ecr.aws/q9t5s3a7"
+    REPO: "vllm-ci-test-repo"
+    VLLM_TEST_DEVICE: "xpu"
+  source_file_dependencies:
+    - vllm/v1/engine/
+    - tests/v1/engine/
+  commands:
+    - >-
+      bash .buildkite/scripts/hardware_ci/run-intel-test.sh
+      'cd tests &&
+      pytest -v -s v1/engine --ignore v1/engine/test_preprocess_error_handling.py'

.buildkite/intel_jobs/kernels_intel.yaml

@@ -0,0 +1,21 @@
+group: Kernels Intel
+depends_on: 
+  - image-build-xpu
+steps:
+- label: vLLM IR Tests
+  timeout_in_minutes: 30
+  device: intel_gpu
+  no_plugin: true
+  working_dir: "."
+  env:
+    REGISTRY: "public.ecr.aws/q9t5s3a7"
+    REPO: "vllm-ci-test-repo"
+    VLLM_TEST_DEVICE: "xpu"
+  source_file_dependencies:
+    - vllm/ir
+    - vllm/kernels
+  commands:
+    - >-
+      bash .buildkite/scripts/hardware_ci/run-intel-test.sh
+      'cd tests &&
+      pytest -v -s kernels/ir'

.buildkite/intel_jobs/lora_intel.yaml

@@ -0,0 +1,129 @@
+group: LoRA Intel
+depends_on:
+  - image-build-xpu
+steps:
+- label: LoRA Runtime + Utils
+  timeout_in_minutes: 45
+  device: intel_gpu
+  no_plugin: true
+  working_dir: "."
+  env:
+    REGISTRY: "public.ecr.aws/q9t5s3a7"
+    REPO: "vllm-ci-test-repo"
+    VLLM_TEST_DEVICE: "xpu"
+  source_file_dependencies:
+    - vllm/lora
+    - tests/lora
+  commands:
+    - >-
+      bash .buildkite/scripts/hardware_ci/run-intel-test.sh
+      'cd tests &&
+      pytest -v -s lora/test_layers.py &&
+      pytest -v -s lora/test_lora_checkpoints.py &&
+      (pytest -v -s lora/test_lora_functions.py --deselect="tests/lora/test_lora_functions.py::test_lora_functions_sync" --deselect="tests/lora/test_lora_functions.py::test_lora_functions_async" || true) &&
+      pytest -v -s lora/test_lora_huggingface.py &&
+      pytest -v -s lora/test_lora_manager.py &&
+      pytest -v -s lora/test_lora_utils.py &&
+      pytest -v -s lora/test_peft_helper.py &&
+      pytest -v -s lora/test_resolver.py &&
+      pytest -v -s lora/test_utils.py &&
+      (pytest -v -s lora/test_add_lora.py --deselect="tests/lora/test_add_lora.py::test_add_lora" || true) &&
+      (pytest -v -s lora/test_worker.py --deselect="tests/lora/test_worker.py::test_worker_apply_lora" || true)'
+
+- label: LoRA Fused/MoE Kernels
+  timeout_in_minutes: 45
+  device: intel_gpu
+  no_plugin: true
+  working_dir: "."
+  env:
+    REGISTRY: "public.ecr.aws/q9t5s3a7"
+    REPO: "vllm-ci-test-repo"
+    VLLM_TEST_DEVICE: "xpu"
+  source_file_dependencies:
+    - vllm/lora
+    - tests/lora
+  commands:
+    - >-
+      bash .buildkite/scripts/hardware_ci/run-intel-test.sh
+      'cd tests &&
+      pytest -v -s lora/test_fused_moe_lora_kernel.py && 
+      pytest -v -s lora/test_moe_lora_align_sum.py'
+
+- label: LoRA Punica Kernels
+  timeout_in_minutes: 45
+  device: intel_gpu
+  no_plugin: true
+  working_dir: "."
+  env:
+    REGISTRY: "public.ecr.aws/q9t5s3a7"
+    REPO: "vllm-ci-test-repo"
+    VLLM_TEST_DEVICE: "xpu"
+  source_file_dependencies:
+    - vllm/lora
+    - tests/lora
+  commands:
+    - >-
+      bash .buildkite/scripts/hardware_ci/run-intel-test.sh
+      'cd tests &&
+      set -o pipefail &&
+      pytest -v -s lora/test_punica_ops.py --deselect="tests/lora/test_punica_ops.py::test_kernels[shrink-0-xpu:0-dtype0-2-2049-64-32-32]" --deselect="tests/lora/test_punica_ops.py::test_kernels_hidden_size[expand-0-xpu:0-dtype1-2-64000-32-4-4]" --deselect="tests/lora/test_punica_ops.py::test_kernels[shrink-0-xpu:0-dtype0-1-2049-128-1-32]" --deselect="tests/lora/test_punica_ops.py::test_kernels[shrink-0-xpu:0-dtype0-1-2049-256-1-4]" --deselect="tests/lora/test_punica_ops.py::test_kernels[shrink-0-xpu:0-dtype0-1-2049-256-8-4]" --deselect="tests/lora/test_punica_ops.py::test_kernels[expand-0-xpu:0-dtype0-3-2049-128-8-16]" --deselect="tests/lora/test_punica_ops.py::test_kernels[shrink-0-xpu:0-dtype0-1-2049-128-8-32]" --deselect="tests/lora/test_punica_ops.py::test_kernels[expand-0-xpu:0-dtype1-1-2049-256-128-32]" --deselect="tests/lora/test_punica_ops.py::test_kernels_hidden_size[shrink-0-xpu:0-dtype0-3-64256-32-4-4]" --deselect="tests/lora/test_punica_ops.py::test_kernels_hidden_size[shrink-0-xpu:0-dtype1-2-29696-32-4-4]" --deselect="tests/lora/test_punica_ops.py::test_kernels_hidden_size[shrink-0-xpu:0-dtype1-3-49408-32-4-4]" --deselect="tests/lora/test_punica_ops.py::test_kernels_hidden_size[shrink-0-xpu:0-dtype0-2-16384-32-4-4]" --deselect="tests/lora/test_punica_ops.py::test_kernels_hidden_size[expand-0-xpu:0-dtype0-2-51328-32-4-4]" --deselect="tests/lora/test_punica_ops.py::test_kernels_hidden_size[expand-0-xpu:0-dtype1-1-102656-32-4-4]"'
+
+- label: LoRA Punica FP8/XPU Ops
+  timeout_in_minutes: 45
+  device: intel_gpu
+  no_plugin: true
+  working_dir: "."
+  env:
+    REGISTRY: "public.ecr.aws/q9t5s3a7"
+    REPO: "vllm-ci-test-repo"
+    VLLM_TEST_DEVICE: "xpu"
+  source_file_dependencies:
+    - vllm/lora
+    - tests/lora
+  commands:
+    - >-
+      bash .buildkite/scripts/hardware_ci/run-intel-test.sh
+      'cd tests &&
+      pytest -v -s lora/test_punica_ops_fp8.py &&
+      pytest -v -s lora/test_punica_xpu_ops.py'
+
+- label: LoRA Models
+  timeout_in_minutes: 45
+  device: intel_gpu
+  no_plugin: true
+  working_dir: "."
+  env:
+    REGISTRY: "public.ecr.aws/q9t5s3a7"
+    REPO: "vllm-ci-test-repo"
+    VLLM_TEST_DEVICE: "xpu"
+  source_file_dependencies:
+    - vllm/lora
+    - tests/lora
+  commands:
+    - >-
+      bash .buildkite/scripts/hardware_ci/run-intel-test.sh
+      'cd tests &&
+      (pytest -v -s lora/test_mixtral.py --deselect="tests/lora/test_mixtral.py::test_mixtral_lora[4]" || true) &&
+      pytest -v -s lora/test_quant_model.py --deselect="tests/lora/test_quant_model.py::test_quant_model_lora[model0]" --deselect="tests/lora/test_quant_model.py::test_quant_model_lora[model1]" --deselect="tests/lora/test_quant_model.py::test_quant_model_tp_equality[model0]" &&
+      pytest -v -s lora/test_qwen35_densemodel_lora.py &&
+      pytest -v -s lora/test_transformers_model.py'
+
+- label: LoRA Multimodal
+  timeout_in_minutes: 45
+  device: intel_gpu
+  no_plugin: true
+  working_dir: "."
+  env:
+    REGISTRY: "public.ecr.aws/q9t5s3a7"
+    REPO: "vllm-ci-test-repo"
+    VLLM_TEST_DEVICE: "xpu"
+  source_file_dependencies:
+    - vllm/lora
+    - tests/lora
+  commands:
+    - >-
+      bash .buildkite/scripts/hardware_ci/run-intel-test.sh
+      'cd tests &&
+      pytest -v -s lora/test_default_mm_loras.py && 
+      (pytest -v -s lora/test_qwen3_unembed.py || true) &&
+      pytest -v -s lora/test_whisper.py'

.buildkite/intel_jobs/misc_intel.yaml

@@ -0,0 +1,55 @@
+group: Miscellaneous Intel
+depends_on:
+  - image-build-xpu
+steps:
+- label: V1 Core + KV + Metrics
+  timeout_in_minutes: 30
+  device: intel_gpu
+  no_plugin: true
+  working_dir: "."
+  env:
+    REGISTRY: "public.ecr.aws/q9t5s3a7"
+    REPO: "vllm-ci-test-repo"
+    VLLM_TEST_DEVICE: "xpu"
+  source_file_dependencies:
+    - vllm/
+    - tests/v1/core
+    - tests/v1/executor
+    - tests/v1/kv_offload
+    - tests/v1/worker
+    - tests/v1/kv_connector/unit
+    - tests/v1/metrics
+    - tests/entrypoints/openai/correctness/test_lmeval.py
+  commands:
+    - >-
+      bash .buildkite/scripts/hardware_ci/run-intel-test.sh
+      'pip install -r requirements/kv_connectors.txt &&
+      export VLLM_WORKER_MULTIPROC_METHOD=spawn &&
+      cd tests &&
+      pytest -v -s v1/executor'
+
+- label: V1 Sample + Logits
+  timeout_in_minutes: 30
+  device: intel_gpu
+  no_plugin: true
+  working_dir: "."
+  env:
+    REGISTRY: "public.ecr.aws/q9t5s3a7"
+    REPO: "vllm-ci-test-repo"
+    VLLM_TEST_DEVICE: "xpu"
+  source_file_dependencies:
+    - vllm/
+    - tests/v1/sample
+    - tests/v1/logits_processors
+    - tests/v1/test_oracle.py
+    - tests/v1/test_request.py
+    - tests/v1/test_outputs.py
+  commands:
+    - >-
+      bash .buildkite/scripts/hardware_ci/run-intel-test.sh
+      'export VLLM_WORKER_MULTIPROC_METHOD=spawn &&
+      cd tests &&
+      pytest -v -s v1/logits_processors &&
+      pytest -v -s v1/test_oracle.py &&
+      pytest -v -s v1/test_request.py &&
+      pytest -v -s v1/test_outputs.py'

.buildkite/release-pipeline.yaml

@@ -1,3 +1,13 @@
+# CUDA architecture lists — following PyTorch RELEASE.md
+# (https://github.com/pytorch/pytorch/blob/main/RELEASE.md)
+# SM86 included for broader Ampere coverage; SM89 for marlin fp8 support
+env:
+  CUDA_ARCH_X86: "7.5 8.0 8.6 8.9 9.0 10.0 12.0+PTX"
+  # aarch64 only architectures: 8.7 for Orin, 11.0 for Thor (since CUDA 13)
+  CUDA_ARCH_AARCH64: "8.0 8.7 8.9 9.0 10.0 11.0 12.0+PTX"
+  CUDA_ARCH_X86_CU129: "7.5 8.0 8.6 8.9 9.0 10.0 12.0"
+  CUDA_ARCH_AARCH64_CU129: "8.0 8.7 8.9 9.0 10.0 12.0"
+
 steps:
   - input: "Provide Release version here"
     id: input-release-version
@@ -14,12 +24,10 @@ steps:
         agents:
           queue: arm64_cpu_queue_release
         commands:
-          # #NOTE: torch_cuda_arch_list is derived from upstream PyTorch build files here:
-          # https://github.com/pytorch/pytorch/blob/main/.ci/aarch64_linux/aarch64_ci_build.sh#L7
-          - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.9.1 --build-arg torch_cuda_arch_list='8.7 8.9 9.0 10.0+PTX 12.0' --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ."
+          - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.9.1 --build-arg torch_cuda_arch_list=\"${CUDA_ARCH_AARCH64_CU129}\" --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ."
           - "mkdir artifacts"
           - "docker run --rm -v $(pwd)/artifacts:/artifacts_host vllm-ci:build-image bash -c 'cp -r dist /artifacts_host && chmod -R a+rw /artifacts_host'"
-          - "bash .buildkite/scripts/upload-nightly-wheels.sh"
+          - "bash .buildkite/scripts/upload-nightly-wheels.sh manylinux_2_31"
         env:
           DOCKER_BUILDKIT: "1"
 
@@ -29,9 +37,7 @@ steps:
         agents:
           queue: arm64_cpu_queue_release
         commands:
-          # #NOTE: torch_cuda_arch_list is derived from upstream PyTorch build files here:
-          # https://github.com/pytorch/pytorch/blob/main/.ci/aarch64_linux/aarch64_ci_build.sh#L7
-          - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=13.0.1 --build-arg torch_cuda_arch_list='8.7 8.9 9.0 10.0+PTX 12.0' --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.1-devel-ubuntu22.04  --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ."
+          - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=13.0.2 --build-arg torch_cuda_arch_list=\"${CUDA_ARCH_AARCH64}\" --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.2-devel-ubuntu22.04  --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ."
           - "mkdir artifacts"
           - "docker run --rm -v $(pwd)/artifacts:/artifacts_host vllm-ci:build-image bash -c 'cp -r dist /artifacts_host && chmod -R a+rw /artifacts_host'"
           - "bash .buildkite/scripts/upload-nightly-wheels.sh manylinux_2_35"
@@ -57,7 +63,7 @@ steps:
         agents:
           queue: cpu_queue_release
         commands:
-          - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.9.1 --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ."
+          - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.9.1 --build-arg torch_cuda_arch_list=\"${CUDA_ARCH_X86_CU129}\" --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ."
           - "mkdir artifacts"
           - "docker run --rm -v $(pwd)/artifacts:/artifacts_host vllm-ci:build-image bash -c 'cp -r dist /artifacts_host && chmod -R a+rw /artifacts_host'"
           - "bash .buildkite/scripts/upload-nightly-wheels.sh manylinux_2_31"
@@ -70,7 +76,7 @@ steps:
         agents:
           queue: cpu_queue_release
         commands:
-          - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=13.0.1 --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.1-devel-ubuntu22.04 --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ."
+          - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=13.0.2 --build-arg torch_cuda_arch_list=\"${CUDA_ARCH_X86}\" --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.2-devel-ubuntu22.04 --tag vllm-ci:build-image --target build --progress plain -f docker/Dockerfile ."
           - "mkdir artifacts"
           - "docker run --rm -v $(pwd)/artifacts:/artifacts_host vllm-ci:build-image bash -c 'cp -r dist /artifacts_host && chmod -R a+rw /artifacts_host'"
           - "bash .buildkite/scripts/upload-nightly-wheels.sh manylinux_2_35"
@@ -108,96 +114,95 @@ steps:
     depends_on: block-build-release-images
     allow_dependency_failure: true
     steps:
-      - label: "Build release image - x86_64 - CUDA 12.9"
+      - label: "Build release image - x86_64 - CUDA 13.0"
         depends_on: ~
         id: build-release-image-x86
         agents:
           queue: cpu_queue_release
         commands:
           - "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"
-          - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.9.1 --build-arg FLASHINFER_AOT_COMPILE=true --build-arg INSTALL_KV_CONNECTORS=true --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m) --target vllm-openai --progress plain -f docker/Dockerfile ."
+          - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=13.0.2 --build-arg torch_cuda_arch_list=\"${CUDA_ARCH_X86}\" --build-arg INSTALL_KV_CONNECTORS=true --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.2-devel-ubuntu22.04 --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m) --target vllm-openai --progress plain -f docker/Dockerfile ."
           - "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)"
           # re-tag to default image tag and push, just in case arm64 build fails
           - "docker tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m) public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT"
           - "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT"
 
-      - label: "Build release image - aarch64 - CUDA 12.9"
+      - label: "Build release image - aarch64 - CUDA 13.0"
         depends_on: ~
         id: build-release-image-arm64
         agents:
           queue: arm64_cpu_queue_release
         commands:
           - "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"
-          - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.9.1 --build-arg FLASHINFER_AOT_COMPILE=true --build-arg torch_cuda_arch_list='8.7 8.9 9.0 10.0+PTX 12.0' --build-arg INSTALL_KV_CONNECTORS=true --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m) --target vllm-openai --progress plain -f docker/Dockerfile ."
+          - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=13.0.2 --build-arg torch_cuda_arch_list=\"${CUDA_ARCH_AARCH64}\" --build-arg INSTALL_KV_CONNECTORS=true --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.2-devel-ubuntu22.04 --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m) --target vllm-openai --progress plain -f docker/Dockerfile ."
           - "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)"
 
-      - label: "Build release image - x86_64 - CUDA 13.0"
+      - label: "Build release image - x86_64 - CUDA 12.9"
         depends_on: ~
-        id: build-release-image-x86-cuda-13-0
+        id: build-release-image-x86-cuda-12-9
         agents:
           queue: cpu_queue_release
         commands:
           - "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"
-          - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=13.0.1 --build-arg INSTALL_KV_CONNECTORS=true --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.1-devel-ubuntu22.04 --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-cu130 --target vllm-openai --progress plain -f docker/Dockerfile ."
-          - "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-cu130"
+          - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.9.1 --build-arg torch_cuda_arch_list=\"${CUDA_ARCH_X86_CU129}\" --build-arg INSTALL_KV_CONNECTORS=true --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-cu129 --target vllm-openai --progress plain -f docker/Dockerfile ."
+          - "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-cu129"
           # re-tag to default image tag and push, just in case arm64 build fails
-          - "docker tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-cu130 public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-cu130"
-          - "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-cu130"
+          - "docker tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-cu129 public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-cu129"
+          - "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-cu129"
 
-      - label: "Build release image - aarch64 - CUDA 13.0"
+      - label: "Build release image - aarch64 - CUDA 12.9"
         depends_on: ~
-        id: build-release-image-arm64-cuda-13-0
+        id: build-release-image-arm64-cuda-12-9
         agents:
           queue: arm64_cpu_queue_release
         commands:
           - "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"
-          # compute capability 12.0 for RTX-50 series / RTX PRO 6000 Blackwell, 12.1 for DGX Spark
-          - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=13.0.1 --build-arg torch_cuda_arch_list='8.7 8.9 9.0 10.0+PTX 12.0 12.1' --build-arg INSTALL_KV_CONNECTORS=true --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.1-devel-ubuntu22.04 --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-cu130 --target vllm-openai --progress plain -f docker/Dockerfile ."
-          - "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-cu130"
+          - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.9.1 --build-arg torch_cuda_arch_list=\"${CUDA_ARCH_AARCH64_CU129}\" --build-arg INSTALL_KV_CONNECTORS=true --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-cu129 --target vllm-openai --progress plain -f docker/Dockerfile ."
+          - "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-cu129"
 
-      - label: "Build release image - x86_64 - CUDA 12.9 - Ubuntu 24.04"
+      - label: "Build release image - x86_64 - CUDA 13.0 - Ubuntu 24.04"
         depends_on: ~
         id: build-release-image-x86-ubuntu2404
         agents:
           queue: cpu_queue_release
         commands:
           - "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"
-          - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.9.1 --build-arg UBUNTU_VERSION=24.04 --build-arg GDRCOPY_OS_VERSION=Ubuntu24_04 --build-arg FLASHINFER_AOT_COMPILE=true --build-arg torch_cuda_arch_list='8.7 8.9 9.0 10.0+PTX 12.0' --build-arg INSTALL_KV_CONNECTORS=true --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-ubuntu2404 --target vllm-openai --progress plain -f docker/Dockerfile ."
+          - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=13.0.2 --build-arg UBUNTU_VERSION=24.04 --build-arg GDRCOPY_OS_VERSION=Ubuntu24_04 --build-arg torch_cuda_arch_list=\"${CUDA_ARCH_X86}\" --build-arg INSTALL_KV_CONNECTORS=true --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.2-devel-ubuntu24.04 --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-ubuntu2404 --target vllm-openai --progress plain -f docker/Dockerfile ."
           - "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-ubuntu2404"
           - "docker tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-ubuntu2404 public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-ubuntu2404"
           - "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-ubuntu2404"
 
-      - label: "Build release image - aarch64 - CUDA 12.9 - Ubuntu 24.04"
+      - label: "Build release image - aarch64 - CUDA 13.0 - Ubuntu 24.04"
         depends_on: ~
         id: build-release-image-arm64-ubuntu2404
         agents:
           queue: arm64_cpu_queue_release
         commands:
           - "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"
-          - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.9.1 --build-arg UBUNTU_VERSION=24.04 --build-arg GDRCOPY_OS_VERSION=Ubuntu24_04 --build-arg FLASHINFER_AOT_COMPILE=true --build-arg torch_cuda_arch_list='8.7 8.9 9.0 10.0+PTX 12.0' --build-arg INSTALL_KV_CONNECTORS=true --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-ubuntu2404 --target vllm-openai --progress plain -f docker/Dockerfile ."
+          - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=13.0.2 --build-arg UBUNTU_VERSION=24.04 --build-arg GDRCOPY_OS_VERSION=Ubuntu24_04 --build-arg torch_cuda_arch_list=\"${CUDA_ARCH_AARCH64}\" --build-arg INSTALL_KV_CONNECTORS=true --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.2-devel-ubuntu24.04 --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-ubuntu2404 --target vllm-openai --progress plain -f docker/Dockerfile ."
           - "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-ubuntu2404"
 
-      - label: "Build release image - x86_64 - CUDA 13.0 - Ubuntu 24.04"
+      - label: "Build release image - x86_64 - CUDA 12.9 - Ubuntu 24.04"
         depends_on: ~
-        id: build-release-image-x86-cuda-13-0-ubuntu2404
+        id: build-release-image-x86-cuda-12-9-ubuntu2404
         agents:
           queue: cpu_queue_release
         commands:
           - "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"
-          - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=13.0.1 --build-arg UBUNTU_VERSION=24.04 --build-arg GDRCOPY_OS_VERSION=Ubuntu24_04 --build-arg FLASHINFER_AOT_COMPILE=true --build-arg torch_cuda_arch_list='8.7 8.9 9.0 10.0+PTX 12.0 12.1' --build-arg INSTALL_KV_CONNECTORS=true --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.1-devel-ubuntu24.04 --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-cu130-ubuntu2404 --target vllm-openai --progress plain -f docker/Dockerfile ."
-          - "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-cu130-ubuntu2404"
-          - "docker tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-cu130-ubuntu2404 public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-cu130-ubuntu2404"
-          - "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-cu130-ubuntu2404"
+          - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.9.1 --build-arg UBUNTU_VERSION=24.04 --build-arg GDRCOPY_OS_VERSION=Ubuntu24_04 --build-arg torch_cuda_arch_list=\"${CUDA_ARCH_X86_CU129}\" --build-arg INSTALL_KV_CONNECTORS=true --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-cu129-ubuntu2404 --target vllm-openai --progress plain -f docker/Dockerfile ."
+          - "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-cu129-ubuntu2404"
+          - "docker tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-cu129-ubuntu2404 public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-cu129-ubuntu2404"
+          - "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-cu129-ubuntu2404"
 
-      - label: "Build release image - aarch64 - CUDA 13.0 - Ubuntu 24.04"
+      - label: "Build release image - aarch64 - CUDA 12.9 - Ubuntu 24.04"
         depends_on: ~
-        id: build-release-image-arm64-cuda-13-0-ubuntu2404
+        id: build-release-image-arm64-cuda-12-9-ubuntu2404
         agents:
           queue: arm64_cpu_queue_release
         commands:
           - "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"
-          - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=13.0.1 --build-arg UBUNTU_VERSION=24.04 --build-arg GDRCOPY_OS_VERSION=Ubuntu24_04 --build-arg FLASHINFER_AOT_COMPILE=true --build-arg torch_cuda_arch_list='8.7 8.9 9.0 10.0+PTX 12.0 12.1' --build-arg INSTALL_KV_CONNECTORS=true --build-arg BUILD_BASE_IMAGE=nvidia/cuda:13.0.1-devel-ubuntu24.04 --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-cu130-ubuntu2404 --target vllm-openai --progress plain -f docker/Dockerfile ."
-          - "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-cu130-ubuntu2404"
+          - "DOCKER_BUILDKIT=1 docker build --build-arg max_jobs=16 --build-arg USE_SCCACHE=1 --build-arg GIT_REPO_CHECK=1 --build-arg CUDA_VERSION=12.9.1 --build-arg UBUNTU_VERSION=24.04 --build-arg GDRCOPY_OS_VERSION=Ubuntu24_04 --build-arg torch_cuda_arch_list=\"${CUDA_ARCH_AARCH64_CU129}\" --build-arg INSTALL_KV_CONNECTORS=true --tag public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-cu129-ubuntu2404 --target vllm-openai --progress plain -f docker/Dockerfile ."
+          - "docker push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-$(uname -m)-cu129-ubuntu2404"
 
       - block: "Build release image for x86_64 CPU"
         key: block-cpu-release-image-build
@@ -238,7 +243,7 @@ steps:
   - group: "Publish release images"
     key: "publish-release-images"
     steps:
-      - label: "Create multi-arch manifest - CUDA 12.9"
+      - label: "Create multi-arch manifest - CUDA 13.0"
         depends_on:
           - build-release-image-x86
           - build-release-image-arm64
@@ -250,7 +255,7 @@ steps:
           - "docker manifest create public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-x86_64 public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-aarch64 --amend"
           - "docker manifest push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT"
 
-      - label: "Annotate release workflow - CUDA 12.9"
+      - label: "Annotate release workflow - CUDA 13.0"
         depends_on:
           - create-multi-arch-manifest
         id: annotate-release-workflow
@@ -259,19 +264,19 @@ steps:
         commands:
           - "bash .buildkite/scripts/annotate-release.sh"
 
-      - label: "Create multi-arch manifest - CUDA 13.0"
+      - label: "Create multi-arch manifest - CUDA 12.9"
         depends_on:
-          - build-release-image-x86-cuda-13-0
-          - build-release-image-arm64-cuda-13-0
-        id: create-multi-arch-manifest-cuda-13-0
+          - build-release-image-x86-cuda-12-9
+          - build-release-image-arm64-cuda-12-9
+        id: create-multi-arch-manifest-cuda-12-9
         agents:
           queue: small_cpu_queue_release
         commands:
           - "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"
-          - "docker manifest create public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-cu130 public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-x86_64-cu130 public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-aarch64-cu130 --amend"
-          - "docker manifest push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-cu130"
+          - "docker manifest create public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-cu129 public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-x86_64-cu129 public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-aarch64-cu129 --amend"
+          - "docker manifest push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-cu129"
 
-      - label: "Create multi-arch manifest - CUDA 12.9 - Ubuntu 24.04"
+      - label: "Create multi-arch manifest - CUDA 13.0 - Ubuntu 24.04"
         depends_on:
           - build-release-image-x86-ubuntu2404
           - build-release-image-arm64-ubuntu2404
@@ -283,17 +288,17 @@ steps:
           - "docker manifest create public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-ubuntu2404 public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-x86_64-ubuntu2404 public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-aarch64-ubuntu2404 --amend"
           - "docker manifest push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-ubuntu2404"
 
-      - label: "Create multi-arch manifest - CUDA 13.0 - Ubuntu 24.04"
+      - label: "Create multi-arch manifest - CUDA 12.9 - Ubuntu 24.04"
         depends_on:
-          - build-release-image-x86-cuda-13-0-ubuntu2404
-          - build-release-image-arm64-cuda-13-0-ubuntu2404
-        id: create-multi-arch-manifest-cuda-13-0-ubuntu2404
+          - build-release-image-x86-cuda-12-9-ubuntu2404
+          - build-release-image-arm64-cuda-12-9-ubuntu2404
+        id: create-multi-arch-manifest-cuda-12-9-ubuntu2404
         agents:
           queue: small_cpu_queue_release
         commands:
           - "aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin public.ecr.aws/q9t5s3a7"
-          - "docker manifest create public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-cu130-ubuntu2404 public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-x86_64-cu130-ubuntu2404 public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-aarch64-cu130-ubuntu2404 --amend"
-          - "docker manifest push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-cu130-ubuntu2404"
+          - "docker manifest create public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-cu129-ubuntu2404 public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-x86_64-cu129-ubuntu2404 public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-aarch64-cu129-ubuntu2404 --amend"
+          - "docker manifest push public.ecr.aws/q9t5s3a7/vllm-release-repo:$BUILDKITE_COMMIT-cu129-ubuntu2404"
 
       - label: "Publish nightly multi-arch image to DockerHub"
         depends_on:
@@ -313,16 +318,16 @@ steps:
           DOCKER_BUILDKIT: "1"
           DOCKERHUB_USERNAME: "vllmbot"
 
-      - label: "Publish nightly multi-arch image to DockerHub - CUDA 13.0"
+      - label: "Publish nightly multi-arch image to DockerHub - CUDA 12.9"
         depends_on:
-          - create-multi-arch-manifest-cuda-13-0
+          - create-multi-arch-manifest-cuda-12-9
         if: build.env("NIGHTLY") == "1"
         agents:
           queue: small_cpu_queue_release
         commands:
-          - "bash .buildkite/scripts/push-nightly-builds.sh cu130"
+          - "bash .buildkite/scripts/push-nightly-builds.sh cu129"
           # Clean up old nightly builds (keep only last 14)
-          - "bash .buildkite/scripts/cleanup-nightly-builds.sh cu130-nightly-"
+          - "bash .buildkite/scripts/cleanup-nightly-builds.sh cu129-nightly-"
         plugins:
           - docker-login#v3.0.0:
               username: vllmbot

.buildkite/scripts/annotate-release.sh

@@ -13,12 +13,12 @@ ROCM_BASE_CACHE_KEY=$(.buildkite/scripts/cache-rocm-base-wheels.sh key)
 buildkite-agent annotate --style 'info' --context 'release-workflow' << EOF
 To download the wheel (by commit):
 \`\`\`
-aws s3 cp s3://vllm-wheels/${BUILDKITE_COMMIT}/vllm-${RELEASE_VERSION}-cp38-abi3-manylinux_2_31_x86_64.whl .
-aws s3 cp s3://vllm-wheels/${BUILDKITE_COMMIT}/vllm-${RELEASE_VERSION}-cp38-abi3-manylinux_2_31_aarch64.whl .
+aws s3 cp s3://vllm-wheels/${BUILDKITE_COMMIT}/vllm-${RELEASE_VERSION}-cp38-abi3-manylinux_2_35_x86_64.whl .
+aws s3 cp s3://vllm-wheels/${BUILDKITE_COMMIT}/vllm-${RELEASE_VERSION}-cp38-abi3-manylinux_2_35_aarch64.whl .
 
-(Optional) For CUDA 13.0:
-aws s3 cp s3://vllm-wheels/${BUILDKITE_COMMIT}/vllm-${RELEASE_VERSION}+cu130-cp38-abi3-manylinux_2_35_x86_64.whl .
-aws s3 cp s3://vllm-wheels/${BUILDKITE_COMMIT}/vllm-${RELEASE_VERSION}+cu130-cp38-abi3-manylinux_2_35_aarch64.whl .
+(Optional) For CUDA 12.9:
+aws s3 cp s3://vllm-wheels/${BUILDKITE_COMMIT}/vllm-${RELEASE_VERSION}+cu129-cp38-abi3-manylinux_2_31_x86_64.whl .
+aws s3 cp s3://vllm-wheels/${BUILDKITE_COMMIT}/vllm-${RELEASE_VERSION}+cu129-cp38-abi3-manylinux_2_31_aarch64.whl .
 
 (Optional) For CPU:
 aws s3 cp s3://vllm-wheels/${BUILDKITE_COMMIT}/vllm-${RELEASE_VERSION}+cpu-cp38-abi3-manylinux_2_35_x86_64.whl .
@@ -33,8 +33,8 @@ To download and upload the image:
 
 docker pull public.ecr.aws/q9t5s3a7/vllm-release-repo:${BUILDKITE_COMMIT}-x86_64
 docker pull public.ecr.aws/q9t5s3a7/vllm-release-repo:${BUILDKITE_COMMIT}-aarch64
-docker pull public.ecr.aws/q9t5s3a7/vllm-release-repo:${BUILDKITE_COMMIT}-x86_64-cu130
-docker pull public.ecr.aws/q9t5s3a7/vllm-release-repo:${BUILDKITE_COMMIT}-aarch64-cu130
+docker pull public.ecr.aws/q9t5s3a7/vllm-release-repo:${BUILDKITE_COMMIT}-x86_64-cu129
+docker pull public.ecr.aws/q9t5s3a7/vllm-release-repo:${BUILDKITE_COMMIT}-aarch64-cu129
 docker pull public.ecr.aws/q9t5s3a7/vllm-release-repo:${ROCM_BASE_CACHE_KEY}-rocm-base
 docker pull public.ecr.aws/q9t5s3a7/vllm-release-repo:${BUILDKITE_COMMIT}-rocm
 docker pull public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo:v${RELEASE_VERSION}
@@ -50,11 +50,11 @@ docker tag vllm/vllm-openai:x86_64 vllm/vllm-openai:v${RELEASE_VERSION}-x86_64
 docker push vllm/vllm-openai:latest-x86_64
 docker push vllm/vllm-openai:v${RELEASE_VERSION}-x86_64
 
-docker tag public.ecr.aws/q9t5s3a7/vllm-release-repo:${BUILDKITE_COMMIT}-x86_64-cu130 vllm/vllm-openai:x86_64-cu130
-docker tag vllm/vllm-openai:x86_64-cu130 vllm/vllm-openai:latest-x86_64-cu130
-docker tag vllm/vllm-openai:x86_64-cu130 vllm/vllm-openai:v${RELEASE_VERSION}-x86_64-cu130
-docker push vllm/vllm-openai:latest-x86_64-cu130
-docker push vllm/vllm-openai:v${RELEASE_VERSION}-x86_64-cu130
+docker tag public.ecr.aws/q9t5s3a7/vllm-release-repo:${BUILDKITE_COMMIT}-x86_64-cu129 vllm/vllm-openai:x86_64-cu129
+docker tag vllm/vllm-openai:x86_64-cu129 vllm/vllm-openai:latest-x86_64-cu129
+docker tag vllm/vllm-openai:x86_64-cu129 vllm/vllm-openai:v${RELEASE_VERSION}-x86_64-cu129
+docker push vllm/vllm-openai:latest-x86_64-cu129
+docker push vllm/vllm-openai:v${RELEASE_VERSION}-x86_64-cu129
 
 docker tag public.ecr.aws/q9t5s3a7/vllm-release-repo:${BUILDKITE_COMMIT}-aarch64 vllm/vllm-openai:aarch64
 docker tag vllm/vllm-openai:aarch64 vllm/vllm-openai:latest-aarch64
@@ -62,11 +62,11 @@ docker tag vllm/vllm-openai:aarch64 vllm/vllm-openai:v${RELEASE_VERSION}-aarch64
 docker push vllm/vllm-openai:latest-aarch64
 docker push vllm/vllm-openai:v${RELEASE_VERSION}-aarch64
 
-docker tag public.ecr.aws/q9t5s3a7/vllm-release-repo:${BUILDKITE_COMMIT}-aarch64-cu130 vllm/vllm-openai:aarch64-cu130
-docker tag vllm/vllm-openai:aarch64-cu130 vllm/vllm-openai:latest-aarch64-cu130
-docker tag vllm/vllm-openai:aarch64-cu130 vllm/vllm-openai:v${RELEASE_VERSION}-aarch64-cu130
-docker push vllm/vllm-openai:latest-aarch64-cu130
-docker push vllm/vllm-openai:v${RELEASE_VERSION}-aarch64-cu130
+docker tag public.ecr.aws/q9t5s3a7/vllm-release-repo:${BUILDKITE_COMMIT}-aarch64-cu129 vllm/vllm-openai:aarch64-cu129
+docker tag vllm/vllm-openai:aarch64-cu129 vllm/vllm-openai:latest-aarch64-cu129
+docker tag vllm/vllm-openai:aarch64-cu129 vllm/vllm-openai:v${RELEASE_VERSION}-aarch64-cu129
+docker push vllm/vllm-openai:latest-aarch64-cu129
+docker push vllm/vllm-openai:v${RELEASE_VERSION}-aarch64-cu129
 
 ## ROCm
 
@@ -104,11 +104,11 @@ docker manifest create vllm/vllm-openai:v${RELEASE_VERSION} vllm/vllm-openai:v${
 docker manifest push vllm/vllm-openai:latest
 docker manifest push vllm/vllm-openai:v${RELEASE_VERSION}
 
-docker manifest rm vllm/vllm-openai:latest-cu130
-docker manifest create vllm/vllm-openai:latest-cu130 vllm/vllm-openai:latest-x86_64-cu130 vllm/vllm-openai:latest-aarch64-cu130
-docker manifest create vllm/vllm-openai:v${RELEASE_VERSION}-cu130 vllm/vllm-openai:v${RELEASE_VERSION}-x86_64-cu130 vllm/vllm-openai:v${RELEASE_VERSION}-aarch64-cu130
-docker manifest push vllm/vllm-openai:latest-cu130
-docker manifest push vllm/vllm-openai:v${RELEASE_VERSION}-cu130
+docker manifest rm vllm/vllm-openai:latest-cu129
+docker manifest create vllm/vllm-openai:latest-cu129 vllm/vllm-openai:latest-x86_64-cu129 vllm/vllm-openai:latest-aarch64-cu129
+docker manifest create vllm/vllm-openai:v${RELEASE_VERSION}-cu129 vllm/vllm-openai:v${RELEASE_VERSION}-x86_64-cu129 vllm/vllm-openai:v${RELEASE_VERSION}-aarch64-cu129
+docker manifest push vllm/vllm-openai:latest-cu129
+docker manifest push vllm/vllm-openai:v${RELEASE_VERSION}-cu129
 
 docker manifest rm vllm/vllm-openai-cpu:latest || true
 docker manifest create vllm/vllm-openai-cpu:latest vllm/vllm-openai-cpu:latest-x86_64 vllm/vllm-openai-cpu:latest-arm64

.buildkite/scripts/check-ray-compatibility.sh

@@ -29,7 +29,7 @@ if python3 -c "import torch; assert torch.version.hip" 2>/dev/null; then
         TORCH_INDEX_URL=""
     fi
 else
-    TORCH_INDEX_URL="https://download.pytorch.org/whl/cu129"
+    TORCH_INDEX_URL="https://download.pytorch.org/whl/cu130"
 fi
 echo ">>> Using PyTorch index: ${TORCH_INDEX_URL:-PyPI default}"
 

.buildkite/scripts/generate-and-upload-nightly-index.sh

@@ -9,7 +9,7 @@ set -ex
 
 BUCKET="vllm-wheels"
 INDICES_OUTPUT_DIR="indices"
-DEFAULT_VARIANT_ALIAS="cu129" # align with vLLM_MAIN_CUDA_VERSION in vllm/envs.py
+DEFAULT_VARIANT_ALIAS="cu130" # align with vLLM_MAIN_CUDA_VERSION in vllm/envs.py
 PYTHON="${PYTHON_PROG:-python3}" # try to read from env var, otherwise use python3
 SUBPATH=$BUILDKITE_COMMIT
 S3_COMMIT_PREFIX="s3://$BUCKET/$SUBPATH/"

.buildkite/scripts/hardware_ci/run-intel-test.sh

@@ -25,22 +25,100 @@ export PYTHONPATH=".."
 ###############################################################################
 
 cleanup_docker() {
+  # Share the same lock with image pull to avoid cleanup/pull races on one node.
+  local docker_lock="/tmp/docker-pull.lock"
+  exec 9>"$docker_lock"
+  flock 9
+
   docker_root=$(docker info -f '{{.DockerRootDir}}')
   if [ -z "$docker_root" ]; then
     echo "Failed to determine Docker root directory." >&2
-    exit 1
+    flock -u 9
+    return 1
   fi
   echo "Docker root directory: $docker_root"
 
   disk_usage=$(df "$docker_root" | tail -1 | awk '{print $5}' | sed 's/%//')
   threshold=70
   if [ "$disk_usage" -gt "$threshold" ]; then
-    echo "Disk usage is above $threshold%. Cleaning up Docker images and volumes..."
-    docker image prune -f
-    docker volume prune -f && docker system prune --force --filter "until=72h" --all
-    echo "Docker images and volumes cleanup completed."
+    echo "Disk usage is above $threshold%. Running aggressive CI image cleanup..."
+    cleanup_old_ci_images "${REGISTRY}/${REPO}" "${image_name}" "${DOCKER_IMAGE_CLEANUP_HOURS:-72}" 1
   else
-    echo "Disk usage is below $threshold%. No cleanup needed."
+    echo "Disk usage is below $threshold%. Checking old CI images anyway."
+    cleanup_old_ci_images "${REGISTRY}/${REPO}" "${image_name}" "${DOCKER_IMAGE_CLEANUP_HOURS:-72}" 0
+  fi
+  echo "Old CI image cleanup completed."
+
+  flock -u 9
+}
+
+cleanup_old_ci_images() {
+  local repo_prefix="$1"
+  local current_image_ref="$2"
+  local ttl_hours="$3"
+  local aggressive_cleanup="$4"
+
+  if [[ -z "$repo_prefix" || "$repo_prefix" == "/" ]]; then
+    echo "Skip old-image cleanup: invalid repo prefix '${repo_prefix}'"
+    return 0
+  fi
+
+  if ! [[ "$ttl_hours" =~ ^[0-9]+$ ]]; then
+    echo "Invalid DOCKER_IMAGE_CLEANUP_HOURS='${ttl_hours}', fallback to 72"
+    ttl_hours=72
+  fi
+
+  local now_epoch cutoff_epoch
+  now_epoch=$(date +%s)
+  cutoff_epoch=$((now_epoch - ttl_hours * 3600))
+
+  local -a used_image_ids
+  mapfile -t used_image_ids < <(docker ps -aq | xargs -r docker inspect --format '{{.Image}}' | sort -u)
+
+  local removed_count=0
+  local examined_count=0
+  declare -A seen_ids=()
+
+  while read -r image_ref image_id; do
+    [[ -z "$image_ref" || -z "$image_id" ]] && continue
+    ((examined_count++))
+
+    # Keep the image this job is going to use.
+    if [[ "$image_ref" == "$current_image_ref" ]]; then
+      continue
+    fi
+
+    # Avoid duplicate deletes when multiple tags point to same image id.
+    if [[ -n "${seen_ids[$image_id]:-}" ]]; then
+      continue
+    fi
+    seen_ids[$image_id]=1
+
+    # Never delete images that are used by any container on this node.
+    if printf '%s\n' "${used_image_ids[@]}" | grep -qx "$image_id"; then
+      continue
+    fi
+
+    local created created_epoch
+    created=$(docker image inspect -f '{{.Created}}' "$image_id" 2>/dev/null || true)
+    [[ -z "$created" ]] && continue
+    created_epoch=$(date -d "$created" +%s 2>/dev/null || true)
+    [[ -z "$created_epoch" ]] && continue
+
+    if (( created_epoch < cutoff_epoch )) || [[ "$aggressive_cleanup" == "1" ]]; then
+      if docker image rm -f "$image_id" >/dev/null 2>&1; then
+        ((removed_count++))
+      fi
+    fi
+  done < <(docker image ls --no-trunc "$repo_prefix" --format '{{.Repository}}:{{.Tag}} {{.ID}}')
+
+  # Also trim old dangling layers; this is safe and does not remove referenced images.
+  docker image prune -f --filter "until=${ttl_hours}h" >/dev/null 2>&1 || true
+
+  if [[ "$aggressive_cleanup" == "1" ]]; then
+    echo "Examined ${examined_count} images under ${repo_prefix}, removed ${removed_count} unused images under disk pressure."
+  else
+    echo "Examined ${examined_count} images under ${repo_prefix}, removed ${removed_count} old images (>${ttl_hours}h)."
   fi
 }
 
@@ -240,7 +318,6 @@ fi
 cleanup_docker
 
 aws ecr-public get-login-password --region us-east-1 | docker login --username AWS --password-stdin "$REGISTRY"
-aws ecr get-login-password --region us-east-1 | docker login --username AWS --password-stdin 936637512419.dkr.ecr.us-east-1.amazonaws.com
 
 # --- Build or pull test image ---
 IMAGE="${IMAGE_TAG_XPU:-${image_name}}"
@@ -266,8 +343,6 @@ fi
 
 remove_docker_container() {
   docker rm -f "${container_name}" || true
-  docker image rm -f "${image_name}" || true
-  docker system prune -f || true
 }
 trap remove_docker_container EXIT
 
@@ -283,6 +358,7 @@ docker run \
     --ipc=host \
     --privileged \
     -v /dev/dri/by-path:/dev/dri/by-path \
+    -v ${HOME}/.cache/huggingface:/root/.cache/huggingface \
     --entrypoint="" \
     -e "HF_TOKEN=${HF_TOKEN:-}" \
     -e "ZE_AFFINITY_MASK=${ZE_AFFINITY_MASK:-}" \

.buildkite/scripts/hardware_ci/run-xpu-test.sh

@@ -12,9 +12,7 @@ docker build -t "${image_name}" -f docker/Dockerfile.xpu .
 
 # Setup cleanup
 remove_docker_container() {
-  docker rm -f "${container_name}" || true;
-  docker image rm -f "${image_name}" || true;
-  docker system prune -f || true;
+  docker rm -f "${container_name}" || true
 }
 trap remove_docker_container EXIT
 

.buildkite/scripts/scheduled_integration_test/qwen30b_a3b_fp8_dp4_async_eplb.sh

@@ -0,0 +1,55 @@
+#!/usr/bin/env bash
+set -euxo pipefail
+
+# args: [THRESHOLD] [NUM_QUESTIONS] [START_PORT]
+THRESHOLD=${1:-0.8}
+NUM_Q=${2:-1319}
+PORT=${3:-8050}
+OUT_DIR=${OUT_DIR:-/tmp/vllm-scheduled}
+mkdir -p "${OUT_DIR}"
+
+wait_for_server() {
+  local port=$1
+  timeout 600 bash -c '
+    until curl -sf "http://127.0.0.1:'"$port"'/health" > /dev/null; do
+      sleep 1
+    done'
+}
+
+MODEL="Qwen/Qwen3-30B-A3B-FP8"
+BACK="allgather_reducescatter"
+
+cleanup() {
+  if [[ -n "${SERVER_PID:-}" ]] && kill -0 "${SERVER_PID}" 2>/dev/null; then
+    kill "${SERVER_PID}" 2>/dev/null || true
+    for _ in {1..20}; do
+      kill -0 "${SERVER_PID}" 2>/dev/null || break
+      sleep 0.5
+    done
+    kill -9 "${SERVER_PID}" 2>/dev/null || true
+  fi
+}
+trap cleanup EXIT
+
+VLLM_DEEP_GEMM_WARMUP=skip \
+vllm serve "$MODEL" \
+--enforce-eager \
+--data-parallel-size 4 \
+--enable-expert-parallel \
+--enable-eplb \
+--all2all-backend "$BACK" \
+--eplb-config '{"window_size":20, "step_interval":100, "use_async":true}' \
+--trust-remote-code \
+--max-model-len 2048 \
+--port "$PORT" &
+SERVER_PID=$!
+wait_for_server "$PORT"
+
+TAG=$(echo "$MODEL" | tr '/: \\n' '_____')
+OUT="${OUT_DIR}/${TAG}_${BACK}.json"
+python3 tests/evals/gsm8k/gsm8k_eval.py --host http://127.0.0.1 --port "$PORT" --num-questions "${NUM_Q}" --save-results "${OUT}"
+python3 - <<PY
+import json; acc=json.load(open('${OUT}'))['accuracy']
+print(f"${MODEL} ${BACK}: accuracy {acc:.3f}")
+assert acc >= ${THRESHOLD}, f"${MODEL} ${BACK} accuracy {acc}"
+PY

.buildkite/test_areas/disaggregated.yaml

@@ -0,0 +1,98 @@
+group: Disaggregated
+depends_on: 
+  - image-build
+steps:
+- label: Distributed NixlConnector PD accuracy (4 GPUs)
+  timeout_in_minutes: 30
+  working_dir: "/vllm-workspace/tests"
+  num_devices: 4
+  source_file_dependencies:
+    - vllm/distributed/kv_transfer/kv_connector/v1/nixl_connector.py
+    - tests/v1/kv_connector/nixl_integration/
+  commands:
+    - uv pip install --system -r /vllm-workspace/requirements/kv_connectors.txt
+    - bash v1/kv_connector/nixl_integration/config_sweep_accuracy_test.sh
+- label: Distributed FlashInfer NixlConnector PD accuracy (4 GPUs)
+  timeout_in_minutes: 30
+  working_dir: "/vllm-workspace/tests"
+  num_devices: 4
+  source_file_dependencies:
+    - vllm/distributed/kv_transfer/kv_connector/v1/nixl_connector.py
+    - tests/v1/kv_connector/nixl_integration/
+  commands:
+    - uv pip install --system -r /vllm-workspace/requirements/kv_connectors.txt
+    - FLASHINFER=1 bash v1/kv_connector/nixl_integration/config_sweep_accuracy_test.sh
+
+- label: DP EP Distributed NixlConnector PD accuracy tests (4 GPUs)
+  timeout_in_minutes: 30
+  working_dir: "/vllm-workspace/tests"
+  num_devices: 4
+  source_file_dependencies:
+    - vllm/distributed/kv_transfer/kv_connector/v1/nixl_connector.py
+    - tests/v1/kv_connector/nixl_integration/
+  commands:
+    - uv pip install --system -r /vllm-workspace/requirements/kv_connectors.txt
+    - DP_EP=1 bash v1/kv_connector/nixl_integration/config_sweep_accuracy_test.sh
+
+- label: CrossLayer KV layout Distributed NixlConnector PD accuracy tests (4 GPUs)
+  timeout_in_minutes: 30
+  working_dir: "/vllm-workspace/tests"
+  num_devices: 4
+  source_file_dependencies:
+    - vllm/distributed/kv_transfer/kv_connector/v1/nixl_connector.py
+    - tests/v1/kv_connector/nixl_integration/
+  commands:
+    - uv pip install --system -r /vllm-workspace/requirements/kv_connectors.txt
+    - CROSS_LAYERS_BLOCKS=True bash v1/kv_connector/nixl_integration/config_sweep_accuracy_test.sh
+
+- label: Hybrid SSM NixlConnector PD accuracy tests (4 GPUs)
+  timeout_in_minutes: 20
+  working_dir: "/vllm-workspace/tests"
+  num_devices: 4
+  source_file_dependencies:
+    - vllm/distributed/kv_transfer/kv_connector/v1/nixl_connector.py
+    - tests/v1/kv_connector/nixl_integration/
+  commands:
+    - uv pip install --system -r /vllm-workspace/requirements/kv_connectors.txt
+    - HYBRID_SSM=1 bash v1/kv_connector/nixl_integration/config_sweep_accuracy_test.sh
+
+- label: MultiConnector (Nixl+Offloading) PD accuracy (2 GPUs)
+  timeout_in_minutes: 30
+  working_dir: "/vllm-workspace/tests"
+  num_devices: 2
+  source_file_dependencies:
+    - vllm/distributed/kv_transfer/kv_connector/v1/nixl_connector.py
+    - vllm/distributed/kv_transfer/kv_connector/v1/multi_connector.py
+    - vllm/distributed/kv_transfer/kv_connector/v1/offloading_connector.py
+    - vllm/distributed/kv_transfer/kv_connector/v1/offloading/
+    - tests/v1/kv_connector/nixl_integration/
+  commands:
+    - uv pip install --system -r /vllm-workspace/requirements/kv_connectors.txt
+    - bash v1/kv_connector/nixl_integration/run_multi_connector_accuracy_test.sh
+
+- label: NixlConnector PD + Spec Decode acceptance (2 GPUs)
+  timeout_in_minutes: 30
+  device: a100
+  working_dir: "/vllm-workspace/tests"
+  num_devices: 2
+  source_file_dependencies:
+    - vllm/distributed/kv_transfer/kv_connector/v1/nixl_connector.py
+    - vllm/v1/worker/kv_connector_model_runner_mixin.py
+    - tests/v1/kv_connector/nixl_integration/
+  commands:
+    - uv pip install --system -r /vllm-workspace/requirements/kv_connectors.txt
+    - bash v1/kv_connector/nixl_integration/spec_decode_acceptance_test.sh
+
+- label: MultiConnector (Nixl+Offloading) PD edge cases (2 GPUs)
+  timeout_in_minutes: 30
+  working_dir: "/vllm-workspace/tests"
+  num_devices: 2
+  source_file_dependencies:
+    - vllm/distributed/kv_transfer/kv_connector/v1/nixl_connector.py
+    - vllm/distributed/kv_transfer/kv_connector/v1/multi_connector.py
+    - vllm/distributed/kv_transfer/kv_connector/v1/offloading_connector.py
+    - vllm/distributed/kv_transfer/kv_connector/v1/offloading/
+    - tests/v1/kv_connector/nixl_integration/
+  commands:
+    - uv pip install --system -r /vllm-workspace/requirements/kv_connectors.txt
+    - bash v1/kv_connector/nixl_integration/run_multi_connector_edge_case_test.sh

.buildkite/test_areas/distributed.yaml

@@ -88,9 +88,8 @@ steps:
   - vllm/distributed/
   - tests/distributed/test_torchrun_example.py
   - tests/distributed/test_torchrun_example_moe.py
-  - examples/offline_inference/rlhf_colocate.py
   - examples/rl/
-  - tests/examples/offline_inference/data_parallel.py
+  - tests/examples/features/data_parallel/data_parallel_offline.py
   commands:
   # https://github.com/NVIDIA/nccl/issues/1838
   - export NCCL_CUMEM_HOST_ENABLE=0
@@ -107,7 +106,7 @@ steps:
   # test with torchrun tp=2 and dp=2 with ep
   - TP_SIZE=2 DP_SIZE=2 ENABLE_EP=1 torchrun --nproc-per-node=4 tests/distributed/test_torchrun_example_moe.py
   # test with internal dp
-  - python3 examples/offline_inference/data_parallel.py --enforce-eager
+  - python3 examples/features/data_parallel/data_parallel_offline.py --enforce-eager
   # rlhf examples
   - VLLM_ALLOW_INSECURE_SERIALIZATION=1 python3 examples/rl/rlhf_nccl.py
   - VLLM_ALLOW_INSECURE_SERIALIZATION=1 python3 examples/rl/rlhf_ipc.py
@@ -159,7 +158,7 @@ steps:
   num_devices: 8
   working_dir: "/vllm-workspace/tests"
   source_file_dependencies:
-  - examples/offline_inference/torchrun_dp_example.py
+  - examples/features/torchrun/torchrun_dp_example_offline.py
   - vllm/config/parallel.py
   - vllm/distributed/
   - vllm/v1/engine/llm_engine.py
@@ -169,7 +168,7 @@ steps:
   # https://github.com/NVIDIA/nccl/issues/1838
   - export NCCL_CUMEM_HOST_ENABLE=0
   # test with torchrun tp=2 and dp=4 with ep
-  - torchrun --nproc-per-node=8 ../examples/offline_inference/torchrun_dp_example.py --tp-size=2 --pp-size=1 --dp-size=4 --enable-ep
+  - torchrun --nproc-per-node=8 ../examples/features/torchrun/torchrun_dp_example_offline.py --tp-size=2 --pp-size=1 --dp-size=4 --enable-ep
 
 - label: Distributed Tests (4 GPUs)(A100)
   device: a100
@@ -194,7 +193,7 @@ steps:
   commands:
     - pytest -v -s tests/distributed/test_context_parallel.py
     - VLLM_ALLOW_INSECURE_SERIALIZATION=1 python3 examples/rl/rlhf_async_new_apis.py
-    - VLLM_USE_DEEP_GEMM=1 VLLM_LOGGING_LEVEL=DEBUG python3 examples/offline_inference/data_parallel.py --model=Qwen/Qwen1.5-MoE-A2.7B -tp=1 -dp=2 --max-model-len=2048 --all2all-backend=deepep_high_throughput
+    - VLLM_USE_DEEP_GEMM=1 VLLM_LOGGING_LEVEL=DEBUG python3 examples/features/data_parallel/data_parallel_offline.py --model=Qwen/Qwen1.5-MoE-A2.7B -tp=1 -dp=2 --max-model-len=2048 --all2all-backend=deepep_high_throughput
     - pytest -v -s tests/v1/distributed/test_dbo.py
     - VLLM_ALLOW_INSECURE_SERIALIZATION=1 pytest -v -s tests/distributed/test_weight_transfer.py
     - pytest -v -s tests/distributed/test_packed_tensor.py
@@ -222,94 +221,9 @@ steps:
   - vllm/executor/
   - vllm/model_executor/models/
   - tests/distributed/
-  - tests/examples/offline_inference/data_parallel.py
+  - tests/examples/features/data_parallel/data_parallel_offline.py
   commands:
-    - ./.buildkite/scripts/run-multi-node-test.sh /vllm-workspace/tests 2 2 $IMAGE_TAG "VLLM_TEST_SAME_HOST=0 torchrun --nnodes 2 --nproc-per-node=2 --rdzv_backend=c10d --rdzv_endpoint=192.168.10.10 distributed/test_same_node.py | grep 'Same node test passed' && NUM_NODES=2 torchrun --nnodes 2 --nproc-per-node=2 --rdzv_backend=c10d --rdzv_endpoint=192.168.10.10 distributed/test_node_count.py | grep 'Node count test passed' && python3 ../examples/offline_inference/data_parallel.py -dp=2 -tp=1 --dp-num-nodes=2 --dp-node-rank=0 --dp-master-addr=192.168.10.10 --dp-master-port=12345 --enforce-eager --trust-remote-code && VLLM_MULTI_NODE=1 pytest -v -s distributed/test_multi_node_assignment.py && VLLM_MULTI_NODE=1 pytest -v -s distributed/test_pipeline_parallel.py" "VLLM_TEST_SAME_HOST=0 torchrun --nnodes 2 --nproc-per-node=2 --rdzv_backend=c10d --rdzv_endpoint=192.168.10.10 distributed/test_same_node.py | grep 'Same node test passed' && NUM_NODES=2 torchrun --nnodes 2 --nproc-per-node=2 --rdzv_backend=c10d --rdzv_endpoint=192.168.10.10 distributed/test_node_count.py | grep 'Node count test passed' && python3 ../examples/offline_inference/data_parallel.py -dp=2 -tp=1 --dp-num-nodes=2 --dp-node-rank=1 --dp-master-addr=192.168.10.10 --dp-master-port=12345 --enforce-eager --trust-remote-code"
-
-- label: Distributed NixlConnector PD accuracy (4 GPUs)
-  timeout_in_minutes: 30
-  working_dir: "/vllm-workspace/tests"
-  num_devices: 4
-  source_file_dependencies:
-    - vllm/distributed/kv_transfer/kv_connector/v1/nixl_connector.py
-    - tests/v1/kv_connector/nixl_integration/
-  commands:
-    - uv pip install --system -r /vllm-workspace/requirements/kv_connectors.txt
-    - bash v1/kv_connector/nixl_integration/config_sweep_accuracy_test.sh
-
-- label: DP EP Distributed NixlConnector PD accuracy tests (4 GPUs)
-  timeout_in_minutes: 30
-  working_dir: "/vllm-workspace/tests"
-  num_devices: 4
-  source_file_dependencies:
-    - vllm/distributed/kv_transfer/kv_connector/v1/nixl_connector.py
-    - tests/v1/kv_connector/nixl_integration/
-  commands:
-    - uv pip install --system -r /vllm-workspace/requirements/kv_connectors.txt
-    - DP_EP=1 bash v1/kv_connector/nixl_integration/config_sweep_accuracy_test.sh
-
-- label: CrossLayer KV layout Distributed NixlConnector PD accuracy tests (4 GPUs)
-  timeout_in_minutes: 30
-  working_dir: "/vllm-workspace/tests"
-  num_devices: 4
-  source_file_dependencies:
-    - vllm/distributed/kv_transfer/kv_connector/v1/nixl_connector.py
-    - tests/v1/kv_connector/nixl_integration/
-  commands:
-    - uv pip install --system -r /vllm-workspace/requirements/kv_connectors.txt
-    - CROSS_LAYERS_BLOCKS=True bash v1/kv_connector/nixl_integration/config_sweep_accuracy_test.sh
-
-- label: Hyrbid SSM NixlConnector PD accuracy tests (4 GPUs)
-  timeout_in_minutes: 20
-  working_dir: "/vllm-workspace/tests"
-  num_devices: 4
-  source_file_dependencies:
-    - vllm/distributed/kv_transfer/kv_connector/v1/nixl_connector.py
-    - tests/v1/kv_connector/nixl_integration/
-  commands:
-    - uv pip install --system -r /vllm-workspace/requirements/kv_connectors.txt
-    - HYBRID_SSM=1 bash v1/kv_connector/nixl_integration/config_sweep_accuracy_test.sh
-
-- label: MultiConnector (Nixl+Offloading) PD accuracy (2 GPUs)
-  timeout_in_minutes: 30
-  working_dir: "/vllm-workspace/tests"
-  num_devices: 2
-  source_file_dependencies:
-    - vllm/distributed/kv_transfer/kv_connector/v1/nixl_connector.py
-    - vllm/distributed/kv_transfer/kv_connector/v1/multi_connector.py
-    - vllm/distributed/kv_transfer/kv_connector/v1/offloading_connector.py
-    - vllm/distributed/kv_transfer/kv_connector/v1/offloading/
-    - tests/v1/kv_connector/nixl_integration/
-  commands:
-    - uv pip install --system -r /vllm-workspace/requirements/kv_connectors.txt
-    - bash v1/kv_connector/nixl_integration/run_multi_connector_accuracy_test.sh
-
-- label: NixlConnector PD + Spec Decode acceptance (2 GPUs)
-  timeout_in_minutes: 30
-  device: a100
-  working_dir: "/vllm-workspace/tests"
-  num_devices: 2
-  source_file_dependencies:
-    - vllm/distributed/kv_transfer/kv_connector/v1/nixl_connector.py
-    - vllm/v1/worker/kv_connector_model_runner_mixin.py
-    - tests/v1/kv_connector/nixl_integration/
-  commands:
-    - uv pip install --system -r /vllm-workspace/requirements/kv_connectors.txt
-    - bash v1/kv_connector/nixl_integration/spec_decode_acceptance_test.sh
-
-- label: MultiConnector (Nixl+Offloading) PD edge cases (2 GPUs)
-  timeout_in_minutes: 30
-  working_dir: "/vllm-workspace/tests"
-  num_devices: 2
-  source_file_dependencies:
-    - vllm/distributed/kv_transfer/kv_connector/v1/nixl_connector.py
-    - vllm/distributed/kv_transfer/kv_connector/v1/multi_connector.py
-    - vllm/distributed/kv_transfer/kv_connector/v1/offloading_connector.py
-    - vllm/distributed/kv_transfer/kv_connector/v1/offloading/
-    - tests/v1/kv_connector/nixl_integration/
-  commands:
-    - uv pip install --system -r /vllm-workspace/requirements/kv_connectors.txt
-    - bash v1/kv_connector/nixl_integration/run_multi_connector_edge_case_test.sh
+    - ./.buildkite/scripts/run-multi-node-test.sh /vllm-workspace/tests 2 2 $IMAGE_TAG "VLLM_TEST_SAME_HOST=0 torchrun --nnodes 2 --nproc-per-node=2 --rdzv_backend=c10d --rdzv_endpoint=192.168.10.10 distributed/test_same_node.py | grep 'Same node test passed' && NUM_NODES=2 torchrun --nnodes 2 --nproc-per-node=2 --rdzv_backend=c10d --rdzv_endpoint=192.168.10.10 distributed/test_node_count.py | grep 'Node count test passed' && python3 ../examples/features/data_parallel/data_parallel_offline.py -dp=2 -tp=1 --dp-num-nodes=2 --dp-node-rank=0 --dp-master-addr=192.168.10.10 --dp-master-port=12345 --enforce-eager --trust-remote-code && VLLM_MULTI_NODE=1 pytest -v -s distributed/test_multi_node_assignment.py && VLLM_MULTI_NODE=1 pytest -v -s distributed/test_pipeline_parallel.py" "VLLM_TEST_SAME_HOST=0 torchrun --nnodes 2 --nproc-per-node=2 --rdzv_backend=c10d --rdzv_endpoint=192.168.10.10 distributed/test_same_node.py | grep 'Same node test passed' && NUM_NODES=2 torchrun --nnodes 2 --nproc-per-node=2 --rdzv_backend=c10d --rdzv_endpoint=192.168.10.10 distributed/test_node_count.py | grep 'Node count test passed' && python3 ../examples/features/data_parallel/data_parallel_offline.py -dp=2 -tp=1 --dp-num-nodes=2 --dp-node-rank=1 --dp-master-addr=192.168.10.10 --dp-master-port=12345 --enforce-eager --trust-remote-code"
 
 - label: Pipeline + Context Parallelism (4 GPUs)
   timeout_in_minutes: 60

.buildkite/test_areas/e2e_integration.yaml

@@ -29,6 +29,15 @@ steps:
   commands:
   - bash .buildkite/scripts/scheduled_integration_test/qwen30b_a3b_fp8_block_ep_eplb.sh 0.8 200 8020 2 1
 
+- label: Qwen3-30B-A3B-FP8 DP4 Async EPLB Accuracy
+  timeout_in_minutes: 60
+  device: h100
+  optional: true
+  num_devices: 4
+  working_dir: "/vllm-workspace"
+  commands:
+  - bash .buildkite/scripts/scheduled_integration_test/qwen30b_a3b_fp8_dp4_async_eplb.sh 0.8 200 8050
+
 - label: DeepSeek V2-Lite Prefetch Offload Accuracy (H100)
   timeout_in_minutes: 60
   device: h100

.buildkite/test_areas/kernels.yaml

@@ -95,11 +95,13 @@ steps:
   - tests/kernels/moe/test_deepgemm.py
   - tests/kernels/moe/test_batched_deepgemm.py
   - tests/kernels/attention/test_deepgemm_attention.py
+  - tests/quantization/test_cutlass_w4a16.py
   commands:
     - pytest -v -s kernels/quantization/test_block_fp8.py
     - pytest -v -s kernels/moe/test_deepgemm.py
     - pytest -v -s kernels/moe/test_batched_deepgemm.py
     - pytest -v -s kernels/attention/test_deepgemm_attention.py
+    - pytest -v -s quantization/test_cutlass_w4a16.py
 
 - label: Kernels (B200)
   timeout_in_minutes: 30
@@ -141,6 +143,7 @@ steps:
     - pytest -v -s tests/kernels/quantization/test_nvfp4_qutlass.py
     - pytest -v -s tests/kernels/quantization/test_mxfp4_qutlass.py
     - pytest -v -s tests/kernels/moe/test_nvfp4_moe.py
+    - pytest -v -s tests/kernels/moe/test_mxfp4_moe.py
     - pytest -v -s tests/kernels/moe/test_ocp_mx_moe.py
     - pytest -v -s tests/kernels/moe/test_flashinfer.py
     - pytest -v -s tests/kernels/moe/test_flashinfer_moe.py

.buildkite/test_areas/misc.yaml

@@ -113,19 +113,19 @@ steps:
     - python3 basic/offline_inference/embed.py
     - python3 basic/offline_inference/score.py
     # for multi-modal models
-    - python3 offline_inference/audio_language.py --seed 0
-    - python3 offline_inference/vision_language.py --seed 0
-    - python3 offline_inference/vision_language_multi_image.py --seed 0
-    - python3 offline_inference/encoder_decoder_multimodal.py --model-type whisper --seed 0
+    - python3 generate/multimodal/audio_language_offline.py --seed 0
+    - python3 generate/multimodal/vision_language_offline.py --seed 0
+    - python3 generate/multimodal/vision_language_multi_image_offline.py --seed 0
+    - python3 generate/multimodal/encoder_decoder_multimodal_offline.py --model-type whisper --seed 0
      # for pooling models
     - python3 pooling/embed/vision_embedding_offline.py --seed 0
     # for features demo
-    - python3 offline_inference/prefix_caching.py
+    - python3 features/automatic_prefix_caching/prefix_caching_offline.py
     - python3 offline_inference/llm_engine_example.py
     - python3 others/tensorize_vllm_model.py --model facebook/opt-125m serialize --serialized-directory /tmp/ --suffix v1 && python3 others/tensorize_vllm_model.py --model facebook/opt-125m deserialize --path-to-tensors /tmp/vllm/facebook/opt-125m/v1/model.tensors
-    - python3 offline_inference/spec_decode.py --test --method eagle --num_spec_tokens 3 --dataset-name hf --dataset-path philschmid/mt-bench --num-prompts 80 --temp 0 --top-p 1.0 --top-k -1 --tp 1 --enable-chunked-prefill --max-model-len 2048
+    - python3 features/speculative_decoding/spec_decode_offline.py --test --method eagle --num_spec_tokens 3 --dataset-name hf --dataset-path philschmid/mt-bench --num-prompts 80 --temp 0 --top-p 1.0 --top-k -1 --tp 1 --enable-chunked-prefill --max-model-len 2048
     # https://github.com/vllm-project/vllm/pull/26682 uses slightly more memory in PyTorch 2.9+ causing this test to OOM in 1xL4 GPU
-    - python3 offline_inference/spec_decode.py --test --method eagle3 --num_spec_tokens 3 --dataset-name hf --dataset-path philschmid/mt-bench --num-prompts 80 --temp 0 --top-p 1.0 --top-k -1 --tp 1 --enable-chunked-prefill --max-model-len 1536
+    - python3 features/speculative_decoding/spec_decode_offline.py --test --method eagle3 --num_spec_tokens 3 --dataset-name hf --dataset-path philschmid/mt-bench --num-prompts 80 --temp 0 --top-p 1.0 --top-k -1 --tp 1 --enable-chunked-prefill --max-model-len 1536
 
 - label: Metrics, Tracing (2 GPUs)
   timeout_in_minutes: 20

.buildkite/test_areas/model_runner_v2.yaml

@@ -31,8 +31,9 @@ steps:
     - vllm/v1/worker/gpu/
     - vllm/v1/core/sched/
     - vllm/v1/worker/gpu_worker.py
-    - examples/offline_inference/
     - examples/basic/offline_inference/
+    - examples/generate/multimodal/
+    - examples/features/
     - examples/pooling/embed/vision_embedding_offline.py
     - examples/others/tensorize_vllm_model.py
   commands:
@@ -44,19 +45,19 @@ steps:
     #- python3 basic/offline_inference/generate.py --model meta-llama/Llama-2-13b-chat-hf --cpu-offload-gb 10  # TODO
     #- python3 basic/offline_inference/embed.py   # TODO
     # for multi-modal models
-    - python3 offline_inference/audio_language.py --seed 0
-    - python3 offline_inference/vision_language.py --seed 0
-    - python3 offline_inference/vision_language_multi_image.py --seed 0
-    - python3 offline_inference/encoder_decoder_multimodal.py --model-type whisper --seed 0
+    - python3 generate/multimodal/audio_language_offline.py --seed 0
+    - python3 generate/multimodal/vision_language_offline.py --seed 0
+    - python3 generate/multimodal/vision_language_multi_image_offline.py --seed 0
+    - python3 generate/multimodal/encoder_decoder_multimodal_offline.py --model-type whisper --seed 0
     # for pooling models
     - python3 pooling/embed/vision_embedding_offline.py --seed 0
     # for features demo
-    - python3 offline_inference/prefix_caching.py
+    - python3 features/automatic_prefix_caching/prefix_caching_offline.py
     - python3 offline_inference/llm_engine_example.py
     - python3 others/tensorize_vllm_model.py --model facebook/opt-125m serialize --serialized-directory /tmp/ --suffix v1 && python3 others/tensorize_vllm_model.py --model facebook/opt-125m deserialize --path-to-tensors /tmp/vllm/facebook/opt-125m/v1/model.tensors
-    - python3 offline_inference/spec_decode.py --test --method eagle --num_spec_tokens 3 --dataset-name hf --dataset-path philschmid/mt-bench --num-prompts 80 --temp 0 --top-p 1.0 --top-k -1 --tp 1 --enable-chunked-prefill --max-model-len 2048
+    - python3 features/speculative_decoding/spec_decode_offline.py --test --method eagle --num_spec_tokens 3 --dataset-name hf --dataset-path philschmid/mt-bench --num-prompts 80 --temp 0 --top-p 1.0 --top-k -1 --tp 1 --enable-chunked-prefill --max-model-len 2048
     # https://github.com/vllm-project/vllm/pull/26682 uses slightly more memory in PyTorch 2.9+ causing this test to OOM in 1xL4 GPU
-    - python3 offline_inference/spec_decode.py --test --method eagle3 --num_spec_tokens 3 --dataset-name hf --dataset-path philschmid/mt-bench --num-prompts 80 --temp 0 --top-p 1.0 --top-k -1 --tp 1 --enable-chunked-prefill --max-model-len 1536
+    - python3 features/speculative_decoding/spec_decode_offline.py --test --method eagle3 --num_spec_tokens 3 --dataset-name hf --dataset-path philschmid/mt-bench --num-prompts 80 --temp 0 --top-p 1.0 --top-k -1 --tp 1 --enable-chunked-prefill --max-model-len 1536
 
 - label: Model Runner V2 Distributed (2 GPUs)
   timeout_in_minutes: 45

.buildkite/test_areas/models_basic.yaml

@@ -69,9 +69,9 @@ steps:
     - pytest -v -s tests/models/multimodal/processing/
     - pytest -v -s tests/models/multimodal/test_mapping.py
     - python3 examples/basic/offline_inference/chat.py
-    - python3 examples/offline_inference/vision_language.py --model-type qwen2_5_vl
+    - python3 examples/generate/multimodal/vision_language_offline.py --model-type qwen2_5_vl
     # Whisper needs spawn method to avoid deadlock
-    - VLLM_WORKER_MULTIPROC_METHOD=spawn python3 examples/offline_inference/audio_language.py --model-type whisper
+    - VLLM_WORKER_MULTIPROC_METHOD=spawn python3 examples/generate/multimodal/audio_language_offline.py --model-type whisper
 
 - label: Transformers Backward Compatibility Models Test
   working_dir: "/vllm-workspace/"
@@ -83,7 +83,7 @@ steps:
     - pytest -v -s tests/models/test_transformers.py
     - pytest -v -s tests/models/multimodal/processing/
     - pytest -v -s tests/models/multimodal/test_mapping.py
-    - python3 examples/offline_inference/basic/chat.py
-    - python3 examples/offline_inference/vision_language.py --model-type qwen2_5_vl
+    - python3 examples/basic/offline_inference/chat.py
+    - python3 examples/generate/multimodal/vision_language_offline.py --model-type qwen2_5_vl
     # Whisper needs spawn method to avoid deadlock
-    - VLLM_WORKER_MULTIPROC_METHOD=spawn python3 examples/offline_inference/audio_language.py --model-type whisper
+    - VLLM_WORKER_MULTIPROC_METHOD=spawn python3 examples/generate/multimodal/audio_language_offline.py --model-type whisper

.buildkite/test_areas/models_multimodal.yaml

@@ -28,6 +28,7 @@ steps:
     - pip install git+https://github.com/TIGER-AI-Lab/Mantis.git
     - pytest -v -s models/multimodal/generation/test_common.py -m core_model -k "qwen3 or gemma"
     - pytest -v -s models/multimodal/generation/test_qwen2_5_vl.py -m core_model
+    - pytest -v -s models/multimodal/generation/test_vit_cudagraph.py -m core_model
   mirror:
     amd:
       device: mi325_1

.buildkite/test_areas/spec_decode.yaml

@@ -12,6 +12,17 @@ steps:
   commands:
     - pytest -v -s v1/e2e/spec_decode -k "eagle_correctness"
 
+- label: Spec Decode Eagle Nightly B200
+  timeout_in_minutes: 30
+  device: b200
+  optional: true
+  source_file_dependencies:
+    - vllm/v1/spec_decode/
+    - vllm/v1/worker/gpu/spec_decode/
+    - tests/v1/e2e/spec_decode/
+  commands:
+    - pytest -v -s v1/e2e/spec_decode -k "eagle_correctness"
+
 - label: Spec Decode Speculators + MTP
   timeout_in_minutes: 30
   device: h200_18gb
@@ -23,6 +34,18 @@ steps:
   commands:
     - pytest -v -s v1/e2e/spec_decode -k "speculators or mtp_correctness"
 
+- label: Spec Decode Speculators + MTP Nightly B200
+  timeout_in_minutes: 30
+  device: b200
+  optional: true
+  source_file_dependencies:
+    - vllm/v1/spec_decode/
+    - vllm/v1/worker/gpu/spec_decode/
+    - vllm/transformers_utils/configs/speculators/
+    - tests/v1/e2e/spec_decode/
+  commands:
+    - pytest -v -s v1/e2e/spec_decode -k "speculators or mtp_correctness"
+  
 - label: Spec Decode Ngram + Suffix
   timeout_in_minutes: 30
   device: h200_18gb
@@ -43,6 +66,17 @@ steps:
   commands:
     - pytest -v -s v1/e2e/spec_decode -k "draft_model or no_sync or batch_inference"
 
+- label: Spec Decode Draft Model Nightly B200
+  timeout_in_minutes: 30
+  device: b200
+  optional: true
+  source_file_dependencies:
+    - vllm/v1/spec_decode/
+    - vllm/v1/worker/gpu/spec_decode/
+    - tests/v1/e2e/spec_decode/
+  commands:
+    - pytest -v -s v1/e2e/spec_decode -k "draft_model or no_sync or batch_inference"
+
 - label: DFlash Speculators Correctness
   timeout_in_minutes: 30
   device: h100

.github/mergify.yml

@@ -262,7 +262,7 @@ pull_request_rules:
       - files~=^docker/Dockerfile.xpu
       - files~=^\\.buildkite/intel_jobs/
       - files=\.buildkite/ci_config_intel.yaml
-      - files=vllm/model_executor/layers/fused_moe/xpu_fused_moe.py
+      - files=vllm/model_executor/layers/fused_moe/experts/xpu_moe.py
       - files=vllm/model_executor/kernels/linear/mixed_precision/xpu.py
       - files=vllm/model_executor/kernels/linear/mxfp8/xpu.py
       - files=vllm/model_executor/kernels/linear/scaled_mm/xpu.py
@@ -308,8 +308,7 @@ pull_request_rules:
       - files=benchmarks/benchmark_serving_structured_output.py
       - files=benchmarks/run_structured_output_benchmark.sh
       - files=docs/features/structured_outputs.md
-      - files=examples/offline_inference/structured_outputs.py
-      - files=examples/online_serving/structured_outputs/structured_outputs.py
+      - files=^examples/features/structured_outputs/
       - files~=^tests/v1/structured_output/
       - files=tests/entrypoints/llm/test_struct_output_generate.py
       - files~=^vllm/v1/structured_output/
@@ -325,7 +324,7 @@ pull_request_rules:
     - or:
       - files~=^vllm/v1/spec_decode/
       - files~=^tests/v1/spec_decode/
-      - files~=^examples/.*(spec_decode|mlpspeculator|eagle|speculation).*\.py
+      - files=^examples/features/speculative_decoding/
       - files~=^vllm/model_executor/models/.*eagle.*\.py
       - files=vllm/model_executor/models/mlp_speculator.py
       - files~=^vllm/transformers_utils/configs/(eagle|medusa|mlp_speculator)\.py
@@ -389,11 +388,7 @@ pull_request_rules:
       - files~=^tests/entrypoints/anthropic/.*tool.*
       - files~=^vllm/tool_parsers/
       - files=docs/features/tool_calling.md
-      - files~=^examples/tool_chat_*
-      - files=examples/offline_inference/chat_with_tools.py
-      - files=examples/online_serving/openai_chat_completion_client_with_tools_required.py
-      - files=examples/online_serving/openai_chat_completion_tool_calls_with_reasoning.py
-      - files=examples/online_serving/openai_chat_completion_client_with_tools.py
+      - files~=^examples/tool_calling/
   actions:
     label:
       add:

.github/workflows/scripts/build.sh

@@ -14,7 +14,7 @@ $python_executable -m pip install -r requirements/build/cuda.txt -r requirements
 # Limit the number of parallel jobs to avoid OOM
 export MAX_JOBS=1
 # Make sure release wheels are built for the following architectures
-export TORCH_CUDA_ARCH_LIST="7.0 7.5 8.0 8.6 8.9 9.0+PTX"
+export TORCH_CUDA_ARCH_LIST="7.5 8.0 8.6 8.9 9.0 10.0 12.0+PTX"
 
 bash tools/check_repo.sh
 

CMakeLists.txt

@@ -34,10 +34,10 @@ install(CODE "set(CMAKE_INSTALL_LOCAL_ONLY TRUE)" ALL_COMPONENTS)
 # Supported python versions.  These versions will be searched in order, the
 # first match will be selected.  These should be kept in sync with setup.py.
 #
-set(PYTHON_SUPPORTED_VERSIONS "3.10" "3.11" "3.12" "3.13")
+set(PYTHON_SUPPORTED_VERSIONS "3.10" "3.11" "3.12" "3.13" "3.14")
 
 # Supported AMD GPU architectures.
-set(HIP_SUPPORTED_ARCHS "gfx906;gfx908;gfx90a;gfx942;gfx950;gfx1030;gfx1100;gfx1101;gfx1150;gfx1151;gfx1152;gfx1153;gfx1200;gfx1201")
+set(HIP_SUPPORTED_ARCHS "gfx906;gfx908;gfx90a;gfx942;gfx950;gfx1030;gfx1100;gfx1101;gfx1102;gfx1103;gfx1150;gfx1151;gfx1152;gfx1153;gfx1200;gfx1201")
 
 # ROCm installation prefix. Default to /opt/rocm but allow override via
 # -DROCM_PATH=/your/rocm/path when invoking cmake.
@@ -94,12 +94,15 @@ find_package(Torch REQUIRED)
 # This check must happen after find_package(Torch) because that's when CMAKE_CUDA_COMPILER_VERSION gets defined
 if(DEFINED CMAKE_CUDA_COMPILER_VERSION AND
    CMAKE_CUDA_COMPILER_VERSION VERSION_GREATER_EQUAL 13.0)
-  set(CUDA_SUPPORTED_ARCHS "7.5;8.0;8.6;8.7;8.9;9.0;10.0;11.0;12.0;12.1")
+   # starting from CUDA 12.9 and Blackwell (10.0), we use family-specific targets (10.0f, 12.0f, etc)
+   # to support the whole generation without specifying all sub-architectures
+   # see: https://developer.nvidia.com/blog/nvidia-blackwell-and-nvidia-cuda-12-9-introduce-family-specific-architecture-features/
+  set(CUDA_SUPPORTED_ARCHS "7.5;8.0;8.6;8.7;8.9;9.0;10.0;11.0;12.0")
 elseif(DEFINED CMAKE_CUDA_COMPILER_VERSION AND
    CMAKE_CUDA_COMPILER_VERSION VERSION_GREATER_EQUAL 12.8)
-  set(CUDA_SUPPORTED_ARCHS "7.0;7.2;7.5;8.0;8.6;8.7;8.9;9.0;10.0;10.1;12.0;12.1")
+  set(CUDA_SUPPORTED_ARCHS "7.5;8.0;8.6;8.7;8.9;9.0;10.0;10.1;10.3;12.0;12.1")
 else()
-  set(CUDA_SUPPORTED_ARCHS "7.0;7.2;7.5;8.0;8.6;8.7;8.9;9.0")
+  set(CUDA_SUPPORTED_ARCHS "7.0;7.5;8.0;8.6;8.7;8.9;9.0")
 endif()
 
 #
@@ -307,7 +310,9 @@ set(VLLM_EXT_SRC
   "csrc/torch_bindings.cpp")
 
 if(VLLM_GPU_LANG STREQUAL "CUDA")
-  list(APPEND VLLM_EXT_SRC "csrc/minimax_reduce_rms_kernel.cu")
+  list(APPEND VLLM_EXT_SRC
+    "csrc/minimax_reduce_rms_kernel.cu"
+    "csrc/fused_deepseek_v4_qnorm_rope_kv_insert_kernel.cu")
 
   SET(CUTLASS_ENABLE_HEADERS_ONLY ON CACHE BOOL "Enable only the header library")
 
@@ -952,7 +957,9 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
       "csrc/libtorch_stable/quantization/fp4/activation_nvfp4_quant_fusion_kernels.cu"
       "csrc/libtorch_stable/quantization/fp4/nvfp4_experts_quant.cu"
       "csrc/libtorch_stable/quantization/fp4/nvfp4_scaled_mm_kernels.cu"
-      "csrc/libtorch_stable/quantization/fp4/nvfp4_blockwise_moe_kernel.cu")
+      "csrc/libtorch_stable/quantization/fp4/nvfp4_blockwise_moe_kernel.cu"
+      "csrc/libtorch_stable/quantization/fp4/mxfp4_experts_quant.cu"
+      "csrc/libtorch_stable/quantization/fp4/mxfp4_blockwise_moe_kernel.cu")
     set_gencode_flags_for_srcs(
       SRCS "${SRCS}"
       CUDA_ARCHS "${FP4_ARCHS}")
@@ -1046,7 +1053,8 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
   list(APPEND VLLM_MOE_EXT_SRC
     "csrc/moe/moe_wna16.cu"
     "csrc/moe/grouped_topk_kernels.cu"
-    "csrc/moe/router_gemm.cu")
+    "csrc/moe/router_gemm.cu"
+    "csrc/moe/topk_softplus_sqrt_kernels.cu")
 endif()
 
 if(VLLM_GPU_LANG STREQUAL "CUDA")

README.md

@@ -14,7 +14,7 @@ Easy, fast, and cheap LLM serving for everyone
 | <a href="https://docs.vllm.ai"><b>Documentation</b></a> | <a href="https://blog.vllm.ai/"><b>Blog</b></a> | <a href="https://arxiv.org/abs/2309.06180"><b>Paper</b></a> | <a href="https://x.com/vllm_project"><b>Twitter/X</b></a> | <a href="https://discuss.vllm.ai"><b>User Forum</b></a> | <a href="https://slack.vllm.ai"><b>Developer Slack</b></a> |
 </p>
 
-🔥 We have built a vllm website to help you get started with vllm. Please visit [vllm.ai](https://vllm.ai) to learn more.
+🔥 We have built a vLLM website to help you get started with vLLM. Please visit [vllm.ai](https://vllm.ai) to learn more.
 For events, please visit [vllm.ai/events](https://vllm.ai/events) to join us.
 
 ---
@@ -50,7 +50,7 @@ vLLM is flexible and easy to use with:
 - Efficient multi-LoRA support for dense and MoE layers
 - Support for NVIDIA GPUs, AMD GPUs, and x86/ARM/PowerPC CPUs. Additionally, diverse hardware plugins such as Google TPUs, Intel Gaudi, IBM Spyre, Huawei Ascend, Rebellions NPU, Apple Silicon, MetaX GPU, and more.
 
-vLLM seamlessly supports 200+ model architectures on HuggingFace, including:
+vLLM seamlessly supports 200+ model architectures on Hugging Face, including:
 
 - Decoder-only LLMs (e.g., Llama, Qwen, Gemma)
 - Mixture-of-Expert LLMs (e.g., Mixtral, DeepSeek-V3, Qwen-MoE, GPT-OSS)

benchmarks/attention_benchmarks/mla_runner.py

@@ -404,6 +404,7 @@ def _build_attention_metadata(
         query_start_loc=q_start_gpu,
         query_start_loc_cpu=q_start_cpu,
         seq_lens=seq_lens_gpu,
+        seq_lens_cpu_upper_bound=seq_lens_cpu,
         _seq_lens_cpu=seq_lens_cpu,
         _num_computed_tokens_cpu=num_computed_tokens_cpu,
         slot_mapping=slot_mapping,

benchmarks/kernels/benchmark_cutlass_moe_fp8.py

@@ -16,7 +16,7 @@ from vllm.model_executor.layers.fused_moe.all2all_utils import (
     maybe_make_prepare_finalize,
 )
 from vllm.model_executor.layers.fused_moe.config import fp8_w8a8_moe_quant_config
-from vllm.model_executor.layers.fused_moe.cutlass_moe import CutlassExpertsFp8
+from vllm.model_executor.layers.fused_moe.experts.cutlass_moe import CutlassExpertsFp8
 from vllm.model_executor.layers.fused_moe.fused_moe import fused_experts, fused_topk
 from vllm.platforms import current_platform
 from vllm.utils.argparse_utils import FlexibleArgumentParser

benchmarks/kernels/benchmark_cutlass_moe_nvfp4.py

@@ -22,7 +22,7 @@ from vllm.model_executor.layers.fused_moe.config import (
     fp8_w8a8_moe_quant_config,
     nvfp4_moe_quant_config,
 )
-from vllm.model_executor.layers.fused_moe.cutlass_moe import (
+from vllm.model_executor.layers.fused_moe.experts.cutlass_moe import (
     CutlassExpertsFp4,
 )
 from vllm.model_executor.layers.fused_moe.fused_moe import fused_experts, fused_topk

benchmarks/kernels/benchmark_grouped_gemm_cutlass.py

@@ -13,7 +13,7 @@ from vllm.model_executor.layers.fused_moe.all2all_utils import (
     maybe_make_prepare_finalize,
 )
 from vllm.model_executor.layers.fused_moe.config import fp8_w8a8_moe_quant_config
-from vllm.model_executor.layers.fused_moe.cutlass_moe import CutlassExpertsFp8
+from vllm.model_executor.layers.fused_moe.experts.cutlass_moe import CutlassExpertsFp8
 from vllm.model_executor.layers.fused_moe.fused_moe import (
     fused_experts,
     fused_topk,

benchmarks/kernels/benchmark_vit_fp8_attn.py

@@ -0,0 +1,324 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+
+# Benchmarks FP8 vs BF16 ViT attention via FlashInfer cuDNN backend.
+#
+# == Usage Examples ==
+#
+# Benchmark mode (default, FlashInfer CUDAGraph Bench)
+#   python3 benchmark_vit_fp8_attn.py
+#
+# Profile mode (PyTorch profiler, saves TensorBoard traces):
+#   python3 benchmark_vit_fp8_attn.py --profile
+#   python3 benchmark_vit_fp8_attn.py --profile --profile-output-dir ./profile_traces
+#
+# Custom seq_lens:
+#   python3 benchmark_vit_fp8_attn.py --seq-lens 4096 8192 16384
+
+from functools import partial
+
+import numpy as np
+import torch
+from torch.profiler import ProfilerActivity, profile, record_function
+
+from vllm.utils.argparse_utils import FlexibleArgumentParser
+
+# Qwen3-VL defaults
+NUM_HEADS = 16
+HEAD_DIM = 72
+DEFAULT_SEQ_LENS = [2304, 4096, 8192, 16384]
+
+
+def _setup_fp8_attention(num_heads: int, head_dim: int) -> tuple:
+    """Create FP8 and BF16 attention modules + workspace."""
+    from types import SimpleNamespace
+    from unittest.mock import patch
+
+    from vllm.config import VllmConfig, set_current_vllm_config
+    from vllm.config.multimodal import MultiModalConfig
+    from vllm.model_executor.layers.attention.mm_encoder_attention import (
+        MMEncoderAttention,
+        _get_flashinfer_workspace_buffer,
+    )
+    from vllm.v1.attention.backends.registry import AttentionBackendEnum
+
+    old_dtype = torch.get_default_dtype()
+    torch.set_default_dtype(torch.bfloat16)
+
+    backend_patch = patch(
+        "vllm.model_executor.layers.attention.mm_encoder_attention"
+        ".get_vit_attn_backend",
+        return_value=AttentionBackendEnum.FLASHINFER,
+    )
+
+    # FP8 attention
+    mm_config_fp8 = MultiModalConfig(mm_encoder_attn_dtype="fp8")
+    vllm_config_fp8 = VllmConfig()
+    vllm_config_fp8.model_config = SimpleNamespace(multimodal_config=mm_config_fp8)
+    with set_current_vllm_config(vllm_config_fp8), backend_patch:
+        attn_fp8 = MMEncoderAttention(
+            num_heads=num_heads,
+            head_size=head_dim,
+            prefix="visual.blocks.0.attn",
+        ).to("cuda")
+
+    # BF16 attention (no FP8)
+    with set_current_vllm_config(VllmConfig()), backend_patch:
+        attn_bf16 = MMEncoderAttention(
+            num_heads=num_heads,
+            head_size=head_dim,
+            prefix="visual.blocks.0.attn",
+        ).to("cuda")
+
+    torch.set_default_dtype(old_dtype)
+
+    workspace = _get_flashinfer_workspace_buffer()
+    return attn_fp8, attn_bf16, workspace
+
+
+def _build_meta(
+    seq_len: int,
+    num_heads: int,
+    head_dim: int,
+    fp8: bool,
+):
+    """Build cu_seqlens, max_seqlen, sequence_lengths."""
+    from vllm.model_executor.layers.attention.mm_encoder_attention import (
+        MMEncoderAttention,
+    )
+    from vllm.utils.math_utils import round_up
+    from vllm.v1.attention.backends.registry import AttentionBackendEnum
+
+    cu_np = np.array([0, seq_len], dtype=np.int32)
+    fp8_padded = num_heads * round_up(head_dim, 16) if fp8 else None
+
+    seq_lengths = MMEncoderAttention.maybe_compute_seq_lens(
+        AttentionBackendEnum.FLASHINFER, cu_np, torch.device("cuda")
+    )
+    max_seqlen = torch.tensor(
+        MMEncoderAttention.compute_max_seqlen(AttentionBackendEnum.FLASHINFER, cu_np),
+        dtype=torch.int32,
+    )
+    cu_seqlens = MMEncoderAttention.maybe_recompute_cu_seqlens(
+        AttentionBackendEnum.FLASHINFER,
+        cu_np,
+        num_heads * head_dim,
+        1,
+        torch.device("cuda"),
+        fp8_padded_hidden_size=fp8_padded,
+    )
+    return cu_seqlens, max_seqlen, seq_lengths
+
+
+def run_benchmark(
+    seq_lens: list[int],
+    num_heads: int,
+    head_dim: int,
+    method: str,
+):
+    """Benchmark FP8 vs BF16 attention across seq_lens.
+
+    Uses FlashInfer GPU-level timing to measure pure kernel time,
+    excluding CPU launch overhead.
+    """
+    if method == "cupti":
+        from flashinfer.testing import bench_gpu_time_with_cupti as bench_fn
+
+        bench_fn = partial(bench_fn, use_cuda_graph=True, cold_l2_cache=False)
+    elif method == "cudagraph":
+        from flashinfer.testing import (
+            bench_gpu_time_with_cudagraph as bench_fn,
+        )
+
+        bench_fn = partial(bench_fn, cold_l2_cache=False)
+    else:
+        raise ValueError(f"Invalid method: {method}")
+
+    attn_fp8, attn_bf16, workspace = _setup_fp8_attention(num_heads, head_dim)
+
+    print(f"Timing method: {method}")
+    print(f"{'seq_len':>8} {'BF16 (us)':>12} {'FP8 (us)':>12} {'Speedup':>10}")
+    print("-" * 46)
+
+    for seq_len in seq_lens:
+        torch.manual_seed(42)
+
+        q = torch.randn(
+            seq_len,
+            num_heads,
+            head_dim,
+            device="cuda",
+            dtype=torch.bfloat16,
+        )
+        k = torch.randn_like(q)
+        v = torch.randn_like(q)
+
+        cu_fp8, max_s, seq_l = _build_meta(seq_len, num_heads, head_dim, fp8=True)
+        # we can reuse cu_fp8 for cu_bf16 since q, k, and v are contiguous
+        cu_bf16 = cu_fp8.clone()
+
+        def bf16_fn(q=q, k=k, v=v, cu=cu_bf16, ms=max_s, sl=seq_l):
+            attn_bf16._forward_flashinfer(q, k, v, cu, ms, sl)
+
+        def fp8_fn(q=q, k=k, v=v, cu=cu_fp8, ms=max_s, sl=seq_l):
+            attn_fp8._forward_flashinfer(q, k, v, cu, ms, sl)
+
+        # bench_fn returns List[float] of per-iteration times in ms
+        bf16_times = bench_fn(bf16_fn)
+        fp8_times = bench_fn(fp8_fn)
+
+        bf16_us = np.median(bf16_times) * 1e3  # ms -> us
+        fp8_us = np.median(fp8_times) * 1e3
+        speedup = bf16_us / fp8_us if fp8_us > 0 else float("inf")
+
+        print(f"{seq_len:>8} {bf16_us:>12.1f} {fp8_us:>12.1f} {speedup:>9.2f}x")
+
+
+def _make_trace_handler(output_dir: str, worker_name: str, label: str):
+    """Create a trace handler that saves to TensorBoard and prints summary."""
+
+    def handler(prof):
+        torch.profiler.tensorboard_trace_handler(output_dir, worker_name)(prof)
+        print(f"\n{'=' * 80}")
+        print(label)
+        print(f"{'=' * 80}")
+        print(prof.key_averages().table(sort_by="cuda_time_total", row_limit=20))
+
+    return handler
+
+
+def run_profile(
+    seq_len: int,
+    num_heads: int,
+    head_dim: int,
+    warmup: int,
+    output_dir: str,
+):
+    """Profile FP8 vs BF16 attention with PyTorch profiler."""
+    attn_fp8, attn_bf16, workspace = _setup_fp8_attention(num_heads, head_dim)
+
+    torch.manual_seed(42)
+    q = torch.randn(
+        seq_len,
+        num_heads,
+        head_dim,
+        device="cuda",
+        dtype=torch.bfloat16,
+    )
+    k = torch.randn_like(q)
+    v = torch.randn_like(q)
+
+    cu_fp8, max_s, seq_l = _build_meta(seq_len, num_heads, head_dim, fp8=True)
+    # we can reuse cu_fp8 for cu_bf16 since q, k, and v are contiguous
+    cu_bf16 = cu_fp8.clone()
+
+    sched = torch.profiler.schedule(wait=0, warmup=warmup, active=1)
+
+    # Profile BF16 (warmup handled by profiler schedule)
+    with profile(
+        activities=[ProfilerActivity.CPU, ProfilerActivity.CUDA],
+        schedule=sched,
+        on_trace_ready=_make_trace_handler(
+            output_dir,
+            f"bf16_h{head_dim}_s{seq_len}",
+            f"BF16 Attention (seq_len={seq_len}, heads={num_heads}, "
+            f"head_dim={head_dim})",
+        ),
+    ) as prof_bf16:
+        for _ in range(warmup + 1):
+            with record_function("bf16_attention"):
+                attn_bf16._forward_flashinfer(
+                    q.clone(), k.clone(), v.clone(), cu_bf16, max_s, seq_l
+                )
+                torch.accelerator.synchronize()
+            prof_bf16.step()
+
+    # Profile FP8 (warmup handled by profiler schedule)
+    with profile(
+        activities=[ProfilerActivity.CPU, ProfilerActivity.CUDA],
+        schedule=sched,
+        on_trace_ready=_make_trace_handler(
+            output_dir,
+            f"fp8_h{head_dim}_s{seq_len}",
+            f"FP8 Attention (seq_len={seq_len}, heads={num_heads}, "
+            f"head_dim={head_dim})",
+        ),
+    ) as prof_fp8:
+        for _ in range(warmup + 1):
+            with record_function("fp8_attention"):
+                attn_fp8._forward_flashinfer(
+                    q.clone(), k.clone(), v.clone(), cu_fp8, max_s, seq_l
+                )
+                torch.accelerator.synchronize()
+            prof_fp8.step()
+
+    print(f"\nTensorBoard traces saved to: {output_dir}")
+    print(f"View with: tensorboard --logdir={output_dir}")
+
+
+if __name__ == "__main__":
+    parser = FlexibleArgumentParser(description="Benchmark FP8 vs BF16 ViT attention.")
+    parser.add_argument(
+        "--seq-lens",
+        type=int,
+        nargs="+",
+        default=DEFAULT_SEQ_LENS,
+        help="Sequence lengths to benchmark",
+    )
+    parser.add_argument(
+        "--num-heads",
+        type=int,
+        default=NUM_HEADS,
+    )
+    parser.add_argument(
+        "--head-dim",
+        type=int,
+        default=HEAD_DIM,
+    )
+    parser.add_argument(
+        "--method",
+        choices=["cupti", "cudagraph"],
+        default="cudagraph",
+        help="GPU timing method: cupti (CUPTI kernel timing) or "
+        "cudagraph (CUDA graph capture/replay). Default: cudagraph",
+    )
+    parser.add_argument(
+        "--warmup",
+        type=int,
+        default=10,
+        help="Warmup iterations (profile mode only)",
+    )
+    parser.add_argument(
+        "--profile",
+        action="store_true",
+        help="Run PyTorch profiler instead of benchmark",
+    )
+    parser.add_argument(
+        "--profile-seq-len",
+        type=int,
+        default=8192,
+        help="Sequence length for profiling (default: 8192)",
+    )
+    parser.add_argument(
+        "--profile-output-dir",
+        type=str,
+        default="./profile_traces",
+        help="Output directory for TensorBoard traces (default: ./profile_traces)",
+    )
+    args = parser.parse_args()
+
+    if args.profile:
+        run_profile(
+            args.profile_seq_len,
+            args.num_heads,
+            args.head_dim,
+            args.warmup,
+            args.profile_output_dir,
+        )
+    else:
+        run_benchmark(
+            args.seq_lens,
+            args.num_heads,
+            args.head_dim,
+            args.method,
+        )

benchmarks/kernels/ir/bench_ir_ops.py

@@ -0,0 +1,378 @@
+#!/usr/bin/env python3
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+"""
+Generic benchmark harness for vLLM IR ops.
+
+Usage:
+    python benchmarks/kernels/ir/bench_ir_ops.py
+    python benchmarks/kernels/ir/bench_ir_ops.py --ops rms_norm
+    python benchmarks/kernels/ir/bench_ir_ops.py --ops rms_norm,silu_mul
+    python benchmarks/kernels/ir/bench_ir_ops.py --no-cuda-graph
+    python benchmarks/kernels/ir/bench_ir_ops.py --ops rms_norm --save-path ./results/
+"""
+
+import argparse
+import contextlib
+import csv
+import dataclasses
+import datetime
+import math
+import os
+import subprocess
+import sys
+import tempfile
+
+# Ensure repo root is on sys.path so `benchmarks` is importable as a package.
+_REPO_ROOT = os.path.abspath(os.path.join(os.path.dirname(__file__), "../../.."))
+if _REPO_ROOT not in sys.path:
+    sys.path.insert(0, _REPO_ROOT)
+
+# Suppress noisy C++ warnings from vllm kernel registration (written to fd 2
+# directly by the dynamic linker, so Python-level sys.stderr redirect won't
+# catch them).
+_saved_fd = os.dup(2)
+try:
+    with open(os.devnull, "w") as _devnull:
+        os.dup2(_devnull.fileno(), 2)
+        import torch
+
+        import vllm.kernels  # noqa: E402, F401
+finally:
+    os.dup2(_saved_fd, 2)
+    os.close(_saved_fd)
+
+from tqdm import tqdm  # noqa: E402
+
+from benchmarks.kernels.ir.shapes import SHAPE_CONFIGS  # noqa: E402  # isort: skip
+from vllm.ir.op import IrOp  # noqa: E402
+from vllm.platforms import current_platform  # noqa: E402
+from vllm.triton_utils import triton  # noqa: E402
+
+
+@dataclasses.dataclass(frozen=True)
+class BenchConfig:
+    use_cuda_graph: bool = True
+    warmup: int = 25
+    rep: int = 100
+
+
+def _pkg_version(name: str) -> str:
+    from importlib.metadata import PackageNotFoundError, version
+
+    with contextlib.suppress(PackageNotFoundError):
+        return version(name)
+    return "not installed"
+
+
+_METADATA_LABELS = {
+    "timestamp": "Timestamp",
+    "git_commit": "Git commit",
+    "vllm": "vLLM",
+    "pytorch": "PyTorch",
+    "cuda_runtime": "CUDA runtime",
+    "triton": "Triton",
+    "cutlass": "CUTLASS",
+    "helion": "Helion",
+    "device": "Device",
+    "bench_mode": "Bench mode",
+    "warmup": "Warmup",
+    "rep": "Repetitions",
+}
+
+
+def collect_env_metadata(cfg: BenchConfig) -> dict[str, str]:
+    from vllm.collect_env import get_env_info
+
+    env = get_env_info()
+
+    git_sha = "unknown"
+    with contextlib.suppress(subprocess.CalledProcessError, FileNotFoundError):
+        git_sha = (
+            subprocess.check_output(
+                ["git", "rev-parse", "--short", "HEAD"], stderr=subprocess.DEVNULL
+            )
+            .decode()
+            .strip()
+        )
+
+    device_name = current_platform.get_device_name()
+
+    warmup_note = " ms" if not cfg.use_cuda_graph else " ms (ignored)"
+    rep_note = " replays" if cfg.use_cuda_graph else " ms"
+
+    return {
+        "timestamp": datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
+        "git_commit": git_sha,
+        "vllm": str(env.vllm_version),
+        "pytorch": str(env.torch_version),
+        "cuda_runtime": str(env.cuda_runtime_version),
+        "triton": triton.__version__,
+        "cutlass": _pkg_version("nvidia-cutlass-dsl"),
+        "helion": _pkg_version("helion"),
+        "device": device_name,
+        "bench_mode": "cuda_graph" if cfg.use_cuda_graph else "eager",
+        "warmup": f"{cfg.warmup}{warmup_note}",
+        "rep": f"{cfg.rep}{rep_note}",
+    }
+
+
+def print_metadata(metadata: dict[str, str]):
+    print("=" * 60)
+    for key, val in metadata.items():
+        print(f"{_METADATA_LABELS.get(key, key) + ':':<16}{val}")
+    print("=" * 60)
+
+
+def _clone_args(args: tuple) -> tuple:
+    return tuple(a.clone() if isinstance(a, torch.Tensor) else a for a in args)
+
+
+# TODO(gmagogsfm): When the `maybe_inplace` PR lands, ops marked as
+# inplace=True will mutate bench_args across iterations. Both CUDA graph
+# and eager modes will accumulate drift from repeated in-place mutation.
+# We need to re-clone inputs per iteration for inplace ops.
+def _bench_one(fn, args, cfg: BenchConfig) -> float:
+    bench_args = _clone_args(args)
+    bench_fn = lambda: fn(*bench_args)
+
+    if cfg.use_cuda_graph:
+        ms = triton.testing.do_bench_cudagraph(bench_fn, rep=cfg.rep, quantiles=[0.5])
+    else:
+        ms = triton.testing.do_bench(
+            bench_fn, warmup=cfg.warmup, rep=cfg.rep, quantiles=[0.5]
+        )
+    return ms * 1000
+
+
+# TODO(gmagogsfm): Once compiled native implementation lands (#38775),
+# the benchmark baseline should be the compiled native (what vLLM runs by
+# default) rather than the uncompiled native implementation.
+def collect_timings(
+    op: IrOp, shape_configs: list[dict], cfg: BenchConfig
+) -> tuple[list[str], list[str], dict[str, dict[str, float]]]:
+    def fmt(v) -> str:
+        return str(v).split(".")[-1] if isinstance(v, torch.dtype) else str(v)
+
+    case_names = [
+        "_".join(f"{k}={fmt(v)}" for k, v in kwargs.items()) for kwargs in shape_configs
+    ]
+    providers = [n for n, impl in op.impls.items() if impl.supported]
+
+    results: dict[str, dict[str, float]] = {c: {} for c in case_names}
+    for provider in providers:
+        impl = op.impls[provider]
+        desc = f"{op.name} / {provider}"
+        for case_name, kwargs in tqdm(
+            zip(case_names, shape_configs),
+            desc=desc,
+            total=len(case_names),
+            unit=" cases",
+        ):
+            args = op.generate_inputs(**kwargs)
+            if impl.supports_args(*args):
+                results[case_name][provider] = _bench_one(impl.impl_fn, args, cfg)
+            else:
+                results[case_name][provider] = float("nan")
+
+    return case_names, providers, results
+
+
+def analyze_results(
+    op_name: str,
+    case_names: list[str],
+    providers: list[str],
+    results: dict[str, dict[str, float]],
+) -> tuple[list[dict[str, str]], list[dict[str, str]], list[str]]:
+    native_col = "native"
+    non_native = [p for p in providers if p != native_col]
+
+    header_cols = ["case"]
+    for p in providers:
+        header_cols.append(f"{p} (us)")
+    for p in non_native:
+        header_cols.append(f"{p} speedup")
+
+    detail_rows: list[dict[str, str]] = []
+    speedup_data: dict[str, list[tuple[float, str]]] = {p: [] for p in non_native}
+
+    for case_name in case_names:
+        timings = results[case_name]
+        row: dict[str, str] = {"case": case_name}
+
+        for p in providers:
+            val = timings.get(p, float("nan"))
+            row[f"{p} (us)"] = f"{val:.2f}" if not math.isnan(val) else "n/a"
+
+        native_us = timings.get(native_col, float("nan"))
+        for p in non_native:
+            p_us = timings.get(p, float("nan"))
+            if not math.isnan(native_us) and not math.isnan(p_us) and p_us > 0:
+                speedup = native_us / p_us
+                row[f"{p} speedup"] = f"{speedup:.2f}x"
+                speedup_data[p].append((speedup, case_name))
+            else:
+                row[f"{p} speedup"] = "n/a"
+
+        detail_rows.append(row)
+
+    summary_rows: list[dict[str, str]] = []
+    for p in non_native:
+        entries = speedup_data[p]
+        if not entries:
+            continue
+        speedups = [s for s, _ in entries]
+        geomean = math.exp(sum(math.log(s) for s in speedups) / len(speedups))
+        best_val, best_case = max(entries)
+        worst_val, worst_case = min(entries)
+        wins = sum(1 for s in speedups if s > 1.0)
+        losses = sum(1 for s in speedups if s < 1.0)
+        total = len(speedups)
+
+        print(f"\n{p} vs native ({wins}/{total} faster, {losses}/{total} slower):")
+        print(f"  geomean speedup: {geomean:.2f}x")
+        print(f"  best:            {best_val:.2f}x  ({best_case})")
+        print(f"  worst:           {worst_val:.2f}x  ({worst_case})")
+
+        summary_rows.append(
+            {
+                "op": op_name,
+                "provider": p,
+                "geomean_speedup": f"{geomean:.2f}",
+                "best_speedup": f"{best_val:.2f}",
+                "best_case": best_case,
+                "worst_speedup": f"{worst_val:.2f}",
+                "worst_case": worst_case,
+                "wins": str(wins),
+                "losses": str(losses),
+                "total": str(total),
+            }
+        )
+
+    return detail_rows, summary_rows, header_cols
+
+
+def write_csv(path: str, rows: list[dict[str, str]], fieldnames: list[str]):
+    with open(path, "w", newline="") as f:
+        writer = csv.DictWriter(f, fieldnames=fieldnames)
+        writer.writeheader()
+        writer.writerows(rows)
+
+
+def save_results(
+    save_dir: str,
+    op_name: str,
+    detail_rows: list[dict[str, str]],
+    header_cols: list[str],
+    all_summary_rows: list[dict[str, str]],
+    metadata: dict[str, str],
+):
+    write_csv(
+        os.path.join(save_dir, f"{op_name}_detail.csv"),
+        detail_rows,
+        header_cols,
+    )
+    if all_summary_rows:
+        write_csv(
+            os.path.join(save_dir, "summary.csv"),
+            all_summary_rows,
+            list(all_summary_rows[0].keys()),
+        )
+    write_csv(
+        os.path.join(save_dir, "metadata.csv"),
+        [metadata],
+        list(metadata.keys()),
+    )
+
+
+def parse_args():
+    parser = argparse.ArgumentParser(description="Benchmark vLLM IR ops")
+    parser.add_argument(
+        "--ops",
+        type=str,
+        default=None,
+        help="Comma-separated list of op names to benchmark (substring match)",
+    )
+    parser.add_argument(
+        "--no-cuda-graph",
+        action="store_true",
+        help="Disable CUDA graph; use do_bench with L2 cache flushing instead",
+    )
+    parser.add_argument(
+        "--warmup",
+        type=int,
+        default=25,
+        help="Warmup time in ms (do_bench) or ignored with CUDA graph (default: 25)",
+    )
+    parser.add_argument(
+        "--rep",
+        type=int,
+        default=100,
+        help="Repetition time in ms (do_bench) or number of graph replays "
+        "(do_bench_cudagraph) (default: 100)",
+    )
+    parser.add_argument(
+        "--save-path",
+        type=str,
+        default=None,
+        help="Directory to save results (default: auto-created temp dir)",
+    )
+    return parser.parse_args()
+
+
+def main():
+    args = parse_args()
+    cfg = BenchConfig(
+        use_cuda_graph=not args.no_cuda_graph,
+        warmup=args.warmup,
+        rep=args.rep,
+    )
+
+    torch.set_default_device(current_platform.device_type)
+
+    metadata = collect_env_metadata(cfg)
+    print_metadata(metadata)
+
+    timestamp = datetime.datetime.now().strftime("%Y%m%d_%H%M%S")
+    save_dir = args.save_path or os.path.join(
+        tempfile.gettempdir(), f"vllm_ir_bench_{timestamp}"
+    )
+    os.makedirs(save_dir, exist_ok=True)
+
+    op_filters = [f.strip() for f in args.ops.split(",")] if args.ops else None
+    all_summary_rows: list[dict[str, str]] = []
+
+    for op in IrOp.registry.values():
+        if op_filters and not any(f in op.name for f in op_filters):
+            continue
+        if not op.has_input_generator:
+            print(f"Skipping op '{op.name}': no input generator registered")
+            continue
+        if op.name not in SHAPE_CONFIGS:
+            raise RuntimeError(
+                f"No benchmark shape config for op '{op.name}'. "
+                f"Add it to benchmarks/kernels/ir/shapes.py"
+            )
+
+        case_names, providers, results = collect_timings(
+            op, SHAPE_CONFIGS[op.name], cfg
+        )
+        detail_rows, summary_rows, header_cols = analyze_results(
+            op.name, case_names, providers, results
+        )
+        all_summary_rows.extend(summary_rows)
+
+        save_results(
+            save_dir,
+            op.name,
+            detail_rows,
+            header_cols,
+            all_summary_rows,
+            metadata,
+        )
+
+    print(f"\nResults saved to: {save_dir}")
+
+
+if __name__ == "__main__":
+    main()

benchmarks/kernels/ir/shapes.py

@@ -0,0 +1,29 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+"""
+Shape configurations for IR op benchmarks.
+"""
+
+import torch
+
+NUM_TOKENS = [1, 2, 4, 16, 64, 256, 1024, 4096, 16384]
+COMMON_HIDDEN_SIZES = [
+    2048,  # Llama 3.2 1B, Qwen 3 MoE 30B-A3B, Gemma 3n
+    3072,  # Gemma 7B/9B
+    4096,  # Llama 3 8B, Qwen 3 8B, Mistral 7B
+    5120,  # Llama 4 Scout 17B-16E
+    7168,  # DeepSeek V3
+    8192,  # Llama 3 70B
+    16384,  # Llama 3 405B
+]
+
+# Each entry maps an op name to a list of kwarg dicts that will be passed
+# to that op's registered input generator via op.generate_inputs(**kwargs).
+SHAPE_CONFIGS: dict[str, list[dict]] = {
+    "rms_norm": [
+        {"num_tokens": n, "hidden_size": d, "dtype": dtype}
+        for dtype in [torch.float16, torch.bfloat16, torch.float32]
+        for d in COMMON_HIDDEN_SIZES
+        for n in NUM_TOKENS
+    ],
+}

cmake/cpu_extension.cmake

@@ -161,16 +161,49 @@ elseif (S390_FOUND)
         "-mtune=native")
 elseif (CMAKE_SYSTEM_PROCESSOR MATCHES "riscv64")
     message(STATUS "RISC-V detected")
-    if(RVV_BF16_FOUND)
-        message(STATUS "BF16 extension detected")
-        set(MARCH_FLAGS -march=rv64gcv_zvfh_zfbfmin_zvfbfmin_zvl128b -mrvv-vector-bits=zvl -mabi=lp64d)
-        add_compile_definitions(RISCV_BF16_SUPPORT)
-    elseif (RVV_FP16_FOUND)
-        message(WARNING "BF16 functionality is not available")
-        set(MARCH_FLAGS -march=rv64gcv_zvfh_zvl128b -mrvv-vector-bits=zvl -mabi=lp64d)
+    # VLLM_RVV_VLEN selects the target VLEN. Auto-detected from /proc/cpuinfo
+    # by default; override with -DVLLM_RVV_VLEN=128 or -DVLLM_RVV_VLEN=256.
+    if(NOT DEFINED VLLM_RVV_VLEN)
+        # Auto-detect: find the largest zvl<N>b in /proc/cpuinfo isa line.
+        if(EXISTS /proc/cpuinfo)
+            file(READ /proc/cpuinfo _cpuinfo)
+            set(_best 0)
+            foreach(_n IN ITEMS 128 256 512 1024)
+                if(_cpuinfo MATCHES "zvl${_n}b")
+                    set(_best ${_n})
+                endif()
+            endforeach()
+            if(_best GREATER 0)
+                set(VLLM_RVV_VLEN ${_best})
+            endif()
+        endif()
+        # If auto-detect failed (no /proc/cpuinfo or no zvl<N>b reported)
+        # but the compiler supports RVV, require explicit specification.
+        if(NOT DEFINED VLLM_RVV_VLEN AND (RVV_FP16_FOUND OR RVV_BF16_FOUND))
+            message(FATAL_ERROR
+                "RISC-V RVV is available but VLEN could not be auto-detected. "
+                "Please specify VLEN explicitly:\n"
+                "  -DVLLM_RVV_VLEN=128   (for VLEN=128 hardware)\n"
+                "  -DVLLM_RVV_VLEN=256   (for VLEN=256 hardware, e.g. Spacemit X100)\n"
+                "  -DVLLM_RVV_VLEN=0     (force scalar, no RVV)")
+        endif()
+    endif()
+    if(VLLM_RVV_VLEN AND VLLM_RVV_VLEN GREATER 0)
+        message(STATUS "RISC-V RVV VLEN=${VLLM_RVV_VLEN}")
+        if(RVV_BF16_FOUND)
+            message(STATUS "BF16 extension detected")
+            set(MARCH_FLAGS -march=rv64gcv_zvfh_zfbfmin_zvfbfmin_zvl${VLLM_RVV_VLEN}b -mrvv-vector-bits=zvl -mabi=lp64d)
+            add_compile_definitions(RISCV_BF16_SUPPORT)
+        elseif(RVV_FP16_FOUND)
+            message(WARNING "BF16 functionality is not available")
+            set(MARCH_FLAGS -march=rv64gcv_zvfh_zvl${VLLM_RVV_VLEN}b -mrvv-vector-bits=zvl -mabi=lp64d)
+        else()
+            message(STATUS "compile riscv with scalar (no FP16/BF16)")
+            set(MARCH_FLAGS -march=rv64gc)
+        endif()
     else()
         message(STATUS "compile riscv with scalar")
-        list(APPEND CXX_COMPILE_FLAGS "-march=rv64gc")
+        set(MARCH_FLAGS -march=rv64gc)
     endif()
     list(APPEND CXX_COMPILE_FLAGS ${MARCH_FLAGS})
 else()

cmake/external_projects/deepgemm.cmake

@@ -20,7 +20,7 @@ else()
   FetchContent_Declare(
     deepgemm
     GIT_REPOSITORY https://github.com/deepseek-ai/DeepGEMM.git
-    GIT_TAG 477618cd51baffca09c4b0b87e97c03fe827ef03
+    GIT_TAG 891d57b4db1071624b5c8fa0d1e51cb317fa709f
     GIT_SUBMODULES "third-party/cutlass" "third-party/fmt"
     GIT_PROGRESS TRUE
     CONFIGURE_COMMAND ""
@@ -120,6 +120,11 @@ if(DEEPGEMM_ARCHS)
     COMPONENT _deep_gemm_C
     FILES_MATCHING PATTERN "*.py")
 
+  install(DIRECTORY "${deepgemm_SOURCE_DIR}/deep_gemm/mega/"
+    DESTINATION vllm/third_party/deep_gemm/mega
+    COMPONENT _deep_gemm_C
+    FILES_MATCHING PATTERN "*.py")
+
   # Generate envs.py (normally generated by DeepGEMM's setup.py build step)
   file(WRITE "${CMAKE_CURRENT_BINARY_DIR}/deep_gemm_envs.py"
     "# Pre-installed environment variables\npersistent_envs = dict()\n")

cmake/external_projects/flashmla.cmake

@@ -19,7 +19,7 @@ else()
   FetchContent_Declare(
         flashmla
         GIT_REPOSITORY https://github.com/vllm-project/FlashMLA
-        GIT_TAG 692917b1cda61b93ac9ee2d846ec54e75afe87b1
+        GIT_TAG a6ec2ba7bd0a7dff98b3f4d3e6b52b159c48d78b
         GIT_PROGRESS TRUE
         CONFIGURE_COMMAND ""
         BUILD_COMMAND ""

csrc/activation_kernels.cu

@@ -11,29 +11,74 @@
 namespace vllm {
 
 template <typename scalar_t, scalar_t (*ACT_FN)(const scalar_t&),
-          bool act_first>
+          bool act_first, bool HAS_CLAMP>
 __device__ __forceinline__ scalar_t compute(const scalar_t& x,
-                                            const scalar_t& y) {
-  return act_first ? ACT_FN(x) * y : x * ACT_FN(y);
+                                            const scalar_t& y,
+                                            const float limit) {
+  if constexpr (act_first) {
+    scalar_t gate = x;
+    scalar_t up = y;
+    if constexpr (HAS_CLAMP) {
+      gate = (scalar_t)fminf((float)gate, limit);
+      up = (scalar_t)fmaxf(fminf((float)up, limit), -limit);
+    }
+    return ACT_FN(gate) * up;
+  } else {
+    scalar_t gate = x;
+    scalar_t up = y;
+    if constexpr (HAS_CLAMP) {
+      gate = (scalar_t)fmaxf(fminf((float)gate, limit), -limit);
+      up = (scalar_t)fminf((float)up, limit);
+    }
+    return gate * ACT_FN(up);
+  }
 }
 
 template <typename packed_t, packed_t (*PACKED_ACT_FN)(const packed_t&),
-          bool act_first>
+          bool act_first, bool HAS_CLAMP>
 __device__ __forceinline__ packed_t packed_compute(const packed_t& x,
-                                                   const packed_t& y) {
-  return act_first ? packed_mul(PACKED_ACT_FN(x), y)
-                   : packed_mul(x, PACKED_ACT_FN(y));
+                                                   const packed_t& y,
+                                                   const float limit) {
+  if constexpr (act_first) {
+    packed_t gate = x;
+    packed_t up = y;
+    if constexpr (HAS_CLAMP) {
+      float2 g = cast_to_float2(gate);
+      float2 u = cast_to_float2(up);
+      g.x = fminf(g.x, limit);
+      g.y = fminf(g.y, limit);
+      u.x = fmaxf(fminf(u.x, limit), -limit);
+      u.y = fmaxf(fminf(u.y, limit), -limit);
+      gate = cast_to_packed<packed_t>(g);
+      up = cast_to_packed<packed_t>(u);
+    }
+    return packed_mul(PACKED_ACT_FN(gate), up);
+  } else {
+    packed_t gate = x;
+    packed_t up = y;
+    if constexpr (HAS_CLAMP) {
+      float2 g = cast_to_float2(gate);
+      float2 u = cast_to_float2(up);
+      g.x = fmaxf(fminf(g.x, limit), -limit);
+      g.y = fmaxf(fminf(g.y, limit), -limit);
+      u.x = fminf(u.x, limit);
+      u.y = fminf(u.y, limit);
+      gate = cast_to_packed<packed_t>(g);
+      up = cast_to_packed<packed_t>(u);
+    }
+    return packed_mul(gate, PACKED_ACT_FN(up));
+  }
 }
 
 // Activation and gating kernel template.
 template <typename scalar_t, typename packed_t,
           scalar_t (*ACT_FN)(const scalar_t&),
           packed_t (*PACKED_ACT_FN)(const packed_t&), bool act_first,
-          bool use_vec, bool use_256b = false>
+          bool use_vec, bool HAS_CLAMP, bool use_256b = false>
 __global__ void act_and_mul_kernel(
     scalar_t* __restrict__ out,          // [..., d]
     const scalar_t* __restrict__ input,  // [..., 2, d]
-    const int d) {
+    const int d, const float limit) {
   const scalar_t* x_ptr = input + blockIdx.x * 2 * d;
   const scalar_t* y_ptr = x_ptr + d;
   scalar_t* out_ptr = out + blockIdx.x * d;
@@ -58,8 +103,9 @@ __global__ void act_and_mul_kernel(
       }
 #pragma unroll
       for (int j = 0; j < pvec_t::NUM_ELTS; j++) {
-        x.elts[j] = packed_compute<packed_t, PACKED_ACT_FN, act_first>(
-            x.elts[j], y.elts[j]);
+        x.elts[j] =
+            packed_compute<packed_t, PACKED_ACT_FN, act_first, HAS_CLAMP>(
+                x.elts[j], y.elts[j], limit);
       }
       if constexpr (use_256b) {
         st256(x, &out_vec[i]);
@@ -72,7 +118,8 @@ __global__ void act_and_mul_kernel(
     for (int64_t idx = threadIdx.x; idx < d; idx += blockDim.x) {
       const scalar_t x = VLLM_LDG(&x_ptr[idx]);
       const scalar_t y = VLLM_LDG(&y_ptr[idx]);
-      out_ptr[idx] = compute<scalar_t, ACT_FN, act_first>(x, y);
+      out_ptr[idx] =
+          compute<scalar_t, ACT_FN, act_first, HAS_CLAMP>(x, y, limit);
     }
   }
 }
@@ -151,8 +198,11 @@ packed_gelu_tanh_kernel(const packed_t& val) {
 
 // Launch activation and gating kernel.
 // Use ACT_FIRST (bool) indicating whether to apply the activation function
-// first.
-#define LAUNCH_ACTIVATION_GATE_KERNEL(KERNEL, PACKED_KERNEL, ACT_FIRST)        \
+// first. HAS_CLAMP (bool) enables pre-activation clamping: gate input is
+// clamped (max only) and up input is clamped (both sides) before the
+// activation function is applied.
+#define LAUNCH_ACTIVATION_GATE_KERNEL(KERNEL, PACKED_KERNEL, ACT_FIRST,        \
+                                      HAS_CLAMP, LIMIT)                        \
   auto dtype = input.scalar_type();                                            \
   int d = input.size(-1) / 2;                                                  \
   int64_t num_tokens = input.numel() / input.size(-1);                         \
@@ -177,8 +227,8 @@ packed_gelu_tanh_kernel(const packed_t& val) {
             scalar_t, typename vllm::PackedTypeConverter<scalar_t>::Type,      \
             KERNEL<scalar_t>,                                                  \
             PACKED_KERNEL<typename vllm::PackedTypeConverter<scalar_t>::Type>, \
-            ACT_FIRST, true, true><<<grid, block, 0, stream>>>(                \
-            out.data_ptr<scalar_t>(), input.data_ptr<scalar_t>(), d);          \
+            ACT_FIRST, true, HAS_CLAMP, true><<<grid, block, 0, stream>>>(     \
+            out.data_ptr<scalar_t>(), input.data_ptr<scalar_t>(), d, LIMIT);   \
       });                                                                      \
     } else {                                                                   \
       VLLM_DISPATCH_FLOATING_TYPES(dtype, "act_and_mul_kernel", [&] {          \
@@ -186,8 +236,8 @@ packed_gelu_tanh_kernel(const packed_t& val) {
             scalar_t, typename vllm::PackedTypeConverter<scalar_t>::Type,      \
             KERNEL<scalar_t>,                                                  \
             PACKED_KERNEL<typename vllm::PackedTypeConverter<scalar_t>::Type>, \
-            ACT_FIRST, true, false><<<grid, block, 0, stream>>>(               \
-            out.data_ptr<scalar_t>(), input.data_ptr<scalar_t>(), d);          \
+            ACT_FIRST, true, HAS_CLAMP, false><<<grid, block, 0, stream>>>(    \
+            out.data_ptr<scalar_t>(), input.data_ptr<scalar_t>(), d, LIMIT);   \
       });                                                                      \
     }                                                                          \
   } else {                                                                     \
@@ -197,8 +247,8 @@ packed_gelu_tanh_kernel(const packed_t& val) {
           scalar_t, typename vllm::PackedTypeConverter<scalar_t>::Type,        \
           KERNEL<scalar_t>,                                                    \
           PACKED_KERNEL<typename vllm::PackedTypeConverter<scalar_t>::Type>,   \
-          ACT_FIRST, false><<<grid, block, 0, stream>>>(                       \
-          out.data_ptr<scalar_t>(), input.data_ptr<scalar_t>(), d);            \
+          ACT_FIRST, false, HAS_CLAMP><<<grid, block, 0, stream>>>(            \
+          out.data_ptr<scalar_t>(), input.data_ptr<scalar_t>(), d, LIMIT);     \
     });                                                                        \
   }
 
@@ -206,7 +256,14 @@ void silu_and_mul(torch::Tensor& out,    // [..., d]
                   torch::Tensor& input)  // [..., 2 * d]
 {
   LAUNCH_ACTIVATION_GATE_KERNEL(vllm::silu_kernel, vllm::packed_silu_kernel,
-                                true);
+                                true, false, 0.0f);
+}
+
+void silu_and_mul_clamp(torch::Tensor& out,    // [..., d]
+                        torch::Tensor& input,  // [..., 2 * d]
+                        double limit) {
+  LAUNCH_ACTIVATION_GATE_KERNEL(vllm::silu_kernel, vllm::packed_silu_kernel,
+                                true, true, (float)limit);
 }
 
 void mul_and_silu(torch::Tensor& out,    // [..., d]
@@ -215,21 +272,21 @@ void mul_and_silu(torch::Tensor& out,    // [..., d]
   // The difference between mul_and_silu and silu_and_mul is that mul_and_silu
   // applies the silu to the latter half of the input.
   LAUNCH_ACTIVATION_GATE_KERNEL(vllm::silu_kernel, vllm::packed_silu_kernel,
-                                false);
+                                false, false, 0.0f);
 }
 
 void gelu_and_mul(torch::Tensor& out,    // [..., d]
                   torch::Tensor& input)  // [..., 2 * d]
 {
   LAUNCH_ACTIVATION_GATE_KERNEL(vllm::gelu_kernel, vllm::packed_gelu_kernel,
-                                true);
+                                true, false, 0.0f);
 }
 
 void gelu_tanh_and_mul(torch::Tensor& out,    // [..., d]
                        torch::Tensor& input)  // [..., 2 * d]
 {
-  LAUNCH_ACTIVATION_GATE_KERNEL(vllm::gelu_tanh_kernel,
-                                vllm::packed_gelu_tanh_kernel, true);
+  LAUNCH_ACTIVATION_GATE_KERNEL(
+      vllm::gelu_tanh_kernel, vllm::packed_gelu_tanh_kernel, true, false, 0.0f);
 }
 
 namespace vllm {

csrc/cache_kernels.cu

@@ -599,6 +599,11 @@ __global__ void cp_gather_indexer_k_quant_cache_kernel(
   const int head_idx = (blockIdx.y * blockDim.x + threadIdx.x) * VEC_SIZE;
   // Find batch index within a block
   __shared__ int batch_idx[BLOCK_Y_SIZE];
+  if (threadIdx.x == 0) {
+    batch_idx[threadIdx.y] = -1;
+  }
+  __syncthreads();
+
   for (int iter = 0; iter < cuda_utils::ceil_div(batch_size, int(blockDim.x));
        iter++) {
     int tid = iter * blockDim.x + threadIdx.x;
@@ -611,16 +616,18 @@ __global__ void cp_gather_indexer_k_quant_cache_kernel(
     }
   }
 
-#ifndef USE_ROCM
-  __syncwarp();
-#endif
+  __syncthreads();
 
-  if (head_idx >= head_dim || token_idx >= num_tokens) {
+  // num_tokens may be an allocation upper bound when Python avoids a D2H sync.
+  // Only tokens covered by the exact device-side cu_seq_lens are valid to
+  // gather.
+  const int batch = batch_idx[threadIdx.y];
+  if (head_idx >= head_dim || token_idx >= num_tokens || batch < 0) {
     return;
   }
-  const int inbatch_seq_idx = token_idx - cu_seq_lens[batch_idx[threadIdx.y]];
-  const int block_idx = block_table[batch_idx[threadIdx.y] * num_blocks +
-                                    inbatch_seq_idx / cache_block_size];
+  const int inbatch_seq_idx = token_idx - cu_seq_lens[batch];
+  const int block_idx =
+      block_table[batch * num_blocks + inbatch_seq_idx / cache_block_size];
   const int64_t src_block_offset = block_idx * block_stride;
   const int64_t cache_inblock_offset =
       (inbatch_seq_idx % cache_block_size) * head_dim + head_idx;
@@ -1490,6 +1497,9 @@ void concat_mla_q(torch::Tensor& ql_nope,  // [num_tokens, num_heads, nope_dim]
   TORCH_CHECK(ql_nope.stride(2) == 1, "ql_nope must have stride 1 in dim 2");
   TORCH_CHECK(q_pe.stride(2) == 1, "q_pe must have stride 1 in dim 2");
   TORCH_CHECK(q_out.stride(2) == 1, "q_out must have stride 1 in dim 2");
+  TORCH_CHECK(ql_nope.scalar_type() == at::ScalarType::Half ||
+                  ql_nope.scalar_type() == at::ScalarType::BFloat16,
+              "ql_nope must be float16 or bfloat16 dtype");
 
   if (num_tokens == 0) return;
 
@@ -1501,7 +1511,7 @@ void concat_mla_q(torch::Tensor& ql_nope,  // [num_tokens, num_heads, nope_dim]
   const at::cuda::OptionalCUDAGuard device_guard(device_of(ql_nope));
   const cudaStream_t stream = at::cuda::getCurrentCUDAStream();
 
-  VLLM_DISPATCH_FLOATING_TYPES(ql_nope.scalar_type(), "concat_mla_q", [&] {
+  VLLM_DISPATCH_HALF_TYPES(ql_nope.scalar_type(), "concat_mla_q", [&] {
     vllm::ConcatMLAQKernel<scalar_t, 512><<<grid_size, block_size, 0, stream>>>(
         q_out.data_ptr<scalar_t>(), ql_nope.data_ptr<scalar_t>(),
         q_pe.data_ptr<scalar_t>(), num_tokens, num_heads, q_out.stride(0),

csrc/cpu/cpu_arch_macros.h

@@ -61,8 +61,23 @@
 #endif
 
 #ifdef __aarch64__
-  // Implementation copied from Arm Optimized Routines (expf AdvSIMD)
+  // Implementation of neon_expf copied from Arm Optimized Routines (expf
+  // AdvSIMD)
   // https://github.com/ARM-software/optimized-routines/blob/master/math/aarch64/advsimd/expf.c
+  //
+  // Additional fast exponential intended for cases where outputs will be
+  // downcasted to FP16 / BF16 (e.g. attention softmax). Accurate within 1 ULP
+  // for FP16 Accurate within 1 ULP for BF16 for inputs in [-87.683, 88.376] &
+  // clamps inputs outside this range to 0 / inf. Implementation is similar to
+  // exp_u20, but:
+  // - uses a third degree polynomial approximation for exp(r) instead of a
+  // fifth degree one, with coefficients re-tuned.
+  // - does not split natural log (ln) into high / low parts
+  // - clamps exp(x) to 0 for x < -87.683113f and inf for x > 88.3762589f
+  // exp(x) = 2^n (exp(r))
+  // r = x - n*ln2, with n = round(x/ln2)
+  // exp(r) ~ poly(r) = 1 + r + r^2 * (c3 + c2 * r)
+  // n = round(x / ln2), r = x - n*ln2
   #include <limits>
   #define DEFINE_FAST_EXP                                                      \
     const float32x4_t inv_ln2 = vdupq_n_f32(0x1.715476p+0f);                   \
@@ -106,7 +121,38 @@
                           result.val[2] = neon_expf(vec.reg.val[2]);           \
                           result.val[3] = neon_expf(vec.reg.val[3]);           \
                           return vec_op::FP32Vec16(result);                    \
-                        };
+                        };                                                     \
+    const float32x4_t lower_bound = vdupq_n_f32(-0x1.5ebb82p+6f);              \
+    const float32x4_t upper_bound = vdupq_n_f32(0x1.61814ap+6f);               \
+    constexpr float ln2 = 0x1.62e43p-1f;                                       \
+    constexpr float f_c2 = 0x1.5592ecp-3f;                                     \
+    const float32x4_t f_c3 = vdupq_n_f32(0x1.017d34p-1f);                      \
+    auto neon_expf_f16 = [&](float32x4_t values) __attribute__((               \
+                             always_inline)) {                                 \
+      const uint32x4_t lt_lower = vcltq_f32(values, lower_bound);              \
+      const uint32x4_t gt_upper = vcgtq_f32(values, upper_bound);              \
+      float32x4_t n = vrndaq_f32(vmulq_f32(values, inv_ln2));                  \
+      float32x4_t r = vfmsq_n_f32(values, n, ln2);                             \
+      uint32x4_t e = vshlq_n_u32(vreinterpretq_u32_s32(vcvtq_s32_f32(n)), 23); \
+      float32x4_t r2 = vmulq_f32(r, r);                                        \
+      float32x4_t q = vfmaq_n_f32(f_c3, r, f_c2);                              \
+      float32x4_t s = vaddq_f32(vdupq_n_f32(1.0f), r);                         \
+      float32x4_t p = vfmaq_f32(s, q, r2);                                     \
+      float32x4_t y =                                                          \
+          vreinterpretq_f32_u32(vaddq_u32(vreinterpretq_u32_f32(p), e));       \
+      y = vbslq_f32(lt_lower, vdupq_n_f32(0.0f), y);                           \
+      y = vbslq_f32(gt_upper, vdupq_n_f32(INFINITY), y);                       \
+      return y;                                                                \
+    };                                                                         \
+    auto fast_exp_f16 = [&](const vec_op::FP32Vec16& vec)                      \
+                            __attribute__((always_inline)) {                   \
+                              float32x4x4_t result;                            \
+                              result.val[0] = neon_expf_f16(vec.reg.val[0]);   \
+                              result.val[1] = neon_expf_f16(vec.reg.val[1]);   \
+                              result.val[2] = neon_expf_f16(vec.reg.val[2]);   \
+                              result.val[3] = neon_expf_f16(vec.reg.val[3]);   \
+                              return vec_op::FP32Vec16(result);                \
+                            };
 
 #endif  // __aarch64__
 

csrc/cpu/cpu_attn_impl.hpp

@@ -1152,7 +1152,11 @@ class AttentionMainLoop {
                        bool use_sink) {
 #ifdef DEFINE_FAST_EXP
       DEFINE_FAST_EXP
+      bool constexpr IsReducedPrecision =
+          std::is_same_v<query_t, c10::BFloat16> ||
+          std::is_same_v<query_t, c10::Half>;
 #endif
+
       using prob_buffer_vec_t = typename VecTypeTrait<prob_buffer_t>::vec_t;
       static_assert(sizeof(prob_buffer_t) <= sizeof(logits_buffer_t));
 
@@ -1201,8 +1205,17 @@ class AttentionMainLoop {
             vec = vec - max_vec;
 
             // compute exp
-#ifdef DEFINE_FAST_EXP
-            vec = fast_exp(vec);
+
+#if defined(DEFINE_FAST_EXP)
+  #ifdef __aarch64__
+            if constexpr (IsReducedPrecision) {
+              vec = fast_exp_f16(vec);
+            } else
+  #endif
+            {
+              vec = fast_exp(vec);
+            }
+
             prob_buffer_vec_t output_vec(vec);
             output_vec.save(curr_prob_buffer_iter);
 #else
@@ -1258,7 +1271,11 @@ class AttentionMainLoop {
                        int32_t kv_tile_token_num, float softcap_scale) {
 #ifdef DEFINE_FAST_EXP
       DEFINE_FAST_EXP
+      bool constexpr IsReducedPrecision =
+          std::is_same_v<query_t, c10::BFloat16> ||
+          std::is_same_v<query_t, c10::Half>;
 #endif
+
       float inv_softcap_scale = 1.0 / softcap_scale;
       vec_op::FP32Vec16 softcap_scale_vec(softcap_scale);
       vec_op::FP32Vec16 inv_softcap_scale_vec(inv_softcap_scale);
@@ -1272,8 +1289,15 @@ class AttentionMainLoop {
           vec_op::FP32Vec16 vec(curr_logits_buffer_iter);
           vec = vec * inv_softcap_scale_vec;
 
-#ifdef DEFINE_FAST_EXP
-          vec = fast_exp(vec);
+#if defined(DEFINE_FAST_EXP)
+  #ifdef __aarch64__
+          if constexpr (IsReducedPrecision) {
+            vec = fast_exp_f16(vec);
+          } else
+  #endif
+          {
+            vec = fast_exp(vec);
+          }
           vec_op::FP32Vec16 inv_vec = ones_vec / vec;
           vec = (vec - inv_vec) / (vec + inv_vec);
 #else

csrc/cpu/cpu_types_riscv.hpp

@@ -1,832 +1,25 @@
 #ifndef CPU_TYPES_RISCV_HPP
 #define CPU_TYPES_RISCV_HPP
 
-#include <algorithm>
-#include <cmath>
-#include <cstring>
-#include <iostream>
-#include <limits>
-#include <riscv_vector.h>
-#include <torch/all.h>
+// RISC-V Vector (RVV) CPU type definitions for vLLM.
+//
+// Supports multiple VLENs via compile-time dispatch. The compiler defines
+// __riscv_v_min_vlen from the zvl<N>b extension in -march. The defs header
+// maps VLEN to the correct LMUL suffixes, and the impl header provides
+// VLEN-independent class implementations.
+//
+// To add support for a new VLEN, add the LMUL mapping in
+// cpu_types_riscv_defs.hpp (the impl header needs no changes).
 
-// ============================================================================
-// Vector Register Type Definitions (VLEN=128 bits)
-// ============================================================================
-
-typedef vfloat16m1_t fixed_vfloat16m1_t
-    __attribute__((riscv_rvv_vector_bits(128)));
-typedef vfloat16m2_t fixed_vfloat16m2_t
-    __attribute__((riscv_rvv_vector_bits(256)));
-
-typedef vfloat32m1_t fixed_vfloat32m1_t
-    __attribute__((riscv_rvv_vector_bits(128)));
-typedef vfloat32m2_t fixed_vfloat32m2_t
-    __attribute__((riscv_rvv_vector_bits(256)));
-typedef vfloat32m4_t fixed_vfloat32m4_t
-    __attribute__((riscv_rvv_vector_bits(512)));
-typedef vfloat32m8_t fixed_vfloat32m8_t
-    __attribute__((riscv_rvv_vector_bits(1024)));
-
-typedef vint32m2_t fixed_vint32m2_t __attribute__((riscv_rvv_vector_bits(256)));
-typedef vint32m4_t fixed_vint32m4_t __attribute__((riscv_rvv_vector_bits(512)));
-
-typedef vuint16m1_t fixed_vuint16m1_t
-    __attribute__((riscv_rvv_vector_bits(128)));
-typedef vuint16m2_t fixed_vuint16m2_t
-    __attribute__((riscv_rvv_vector_bits(256)));
-typedef vuint16m4_t fixed_vuint16m4_t
-    __attribute__((riscv_rvv_vector_bits(512)));
-
-#ifdef RISCV_BF16_SUPPORT
-typedef vbfloat16m1_t fixed_vbfloat16m1_t
-    __attribute__((riscv_rvv_vector_bits(128)));
-typedef vbfloat16m2_t fixed_vbfloat16m2_t
-    __attribute__((riscv_rvv_vector_bits(256)));
-typedef vbfloat16m4_t fixed_vbfloat16m4_t
-    __attribute__((riscv_rvv_vector_bits(512)));
+#ifndef __riscv_vector
+  #error "cpu_types_riscv.hpp included in a non-RVV translation unit"
 #endif
 
-namespace vec_op {
-
-#ifdef RISCV_BF16_SUPPORT
-  #define VLLM_DISPATCH_CASE_FLOATING_TYPES(...)         \
-    AT_DISPATCH_CASE(at::ScalarType::Float, __VA_ARGS__) \
-    AT_DISPATCH_CASE(at::ScalarType::Half, __VA_ARGS__)  \
-    AT_DISPATCH_CASE(at::ScalarType::BFloat16, __VA_ARGS__)
-#else
-  #define VLLM_DISPATCH_CASE_FLOATING_TYPES(...)         \
-    AT_DISPATCH_CASE(at::ScalarType::Float, __VA_ARGS__) \
-    AT_DISPATCH_CASE(at::ScalarType::Half, __VA_ARGS__)
+#ifndef __riscv_v_min_vlen
+  #error "compiler did not define __riscv_v_min_vlen; pass -march=...zvl<N>b"
 #endif
 
-#define VLLM_DISPATCH_FLOATING_TYPES(TYPE, NAME, ...) \
-  AT_DISPATCH_SWITCH(TYPE, NAME, VLLM_DISPATCH_CASE_FLOATING_TYPES(__VA_ARGS__))
+#include "cpu_types_riscv_defs.hpp"
+#include "cpu_types_riscv_impl.hpp"
 
-#define FORCE_INLINE __attribute__((always_inline)) inline
-
-namespace {
-template <typename T, T... indexes, typename F>
-constexpr void unroll_loop_item(std::integer_sequence<T, indexes...>, F&& f) {
-  (f(std::integral_constant<T, indexes>{}), ...);
-};
-}  // namespace
-
-template <typename T, T count, typename F,
-          typename = std::enable_if_t<std::is_invocable_v<F, T>>>
-constexpr void unroll_loop(F&& f) {
-  unroll_loop_item(std::make_integer_sequence<T, count>{}, std::forward<F>(f));
-}
-
-template <typename T>
-struct Vec {
-  constexpr static int get_elem_num() { return T::VEC_ELEM_NUM; };
-};
-
-struct FP32Vec8;
-struct FP32Vec16;
-
-// ============================================================================
-// FP16 Implementation
-// ============================================================================
-
-struct FP16Vec8 : public Vec<FP16Vec8> {
-  constexpr static int VEC_ELEM_NUM = 8;
-  fixed_vfloat16m1_t reg;
-
-  explicit FP16Vec8(const void* ptr)
-      : reg(__riscv_vle16_v_f16m1(static_cast<const _Float16*>(ptr),
-                                  VEC_ELEM_NUM)) {};
-
-  explicit FP16Vec8(const FP32Vec8&);
-
-  void save(void* ptr) const {
-    __riscv_vse16_v_f16m1(static_cast<_Float16*>(ptr), reg, VEC_ELEM_NUM);
-  }
-  void save(void* ptr, int elem_num) const {
-    __riscv_vse16_v_f16m1(static_cast<_Float16*>(ptr), reg, elem_num);
-  }
-  void save_strided(void* ptr, ptrdiff_t stride) const {
-    ptrdiff_t byte_stride = stride * sizeof(_Float16);
-    __riscv_vsse16_v_f16m1(static_cast<_Float16*>(ptr), byte_stride, reg,
-                           VEC_ELEM_NUM);
-  }
-};
-
-struct FP16Vec16 : public Vec<FP16Vec16> {
-  constexpr static int VEC_ELEM_NUM = 16;
-  fixed_vfloat16m2_t reg;
-
-  explicit FP16Vec16(const void* ptr)
-      : reg(__riscv_vle16_v_f16m2(static_cast<const _Float16*>(ptr),
-                                  VEC_ELEM_NUM)) {};
-
-  explicit FP16Vec16(const FP32Vec16& vec);
-
-  void save(void* ptr) const {
-    __riscv_vse16_v_f16m2(static_cast<_Float16*>(ptr), reg, VEC_ELEM_NUM);
-  }
-  void save(void* ptr, int elem_num) const {
-    __riscv_vse16_v_f16m2(static_cast<_Float16*>(ptr), reg, elem_num);
-  }
-  void save_strided(void* ptr, ptrdiff_t stride) const {
-    ptrdiff_t byte_stride = stride * sizeof(_Float16);
-    __riscv_vsse16_v_f16m2(static_cast<_Float16*>(ptr), byte_stride, reg,
-                           VEC_ELEM_NUM);
-  }
-};
-
-// ============================================================================
-// BF16 Implementation
-// ============================================================================
-
-#ifdef RISCV_BF16_SUPPORT
-
-FORCE_INLINE fixed_vuint16m1_t bf16_to_u16(fixed_vbfloat16m1_t v) {
-  return __riscv_vreinterpret_v_bf16m1_u16m1(v);
-}
-FORCE_INLINE fixed_vuint16m2_t bf16_to_u16(fixed_vbfloat16m2_t v) {
-  return __riscv_vreinterpret_v_bf16m2_u16m2(v);
-}
-FORCE_INLINE fixed_vuint16m4_t bf16_to_u16(fixed_vbfloat16m4_t v) {
-  return __riscv_vreinterpret_v_bf16m4_u16m4(v);
-}
-
-struct BF16Vec8 : public Vec<BF16Vec8> {
-  constexpr static int VEC_ELEM_NUM = 8;
-  fixed_vbfloat16m1_t reg;
-
-  explicit BF16Vec8(const void* ptr)
-      : reg(__riscv_vreinterpret_v_u16m1_bf16m1(__riscv_vle16_v_u16m1(
-            reinterpret_cast<const uint16_t*>(ptr), VEC_ELEM_NUM))) {};
-
-  explicit BF16Vec8(fixed_vbfloat16m1_t data) : reg(data) {};
-  explicit BF16Vec8(const FP32Vec8&);
-
-  void save(void* ptr) const {
-    __riscv_vse16_v_u16m1(reinterpret_cast<uint16_t*>(ptr), bf16_to_u16(reg),
-                          VEC_ELEM_NUM);
-  }
-  void save(void* ptr, int elem_num) const {
-    __riscv_vse16_v_u16m1(reinterpret_cast<uint16_t*>(ptr), bf16_to_u16(reg),
-                          elem_num);
-  }
-  void save_strided(void* ptr, ptrdiff_t stride) const {
-    ptrdiff_t byte_stride = stride * sizeof(uint16_t);
-    __riscv_vsse16_v_u16m1(reinterpret_cast<uint16_t*>(ptr), byte_stride,
-                           bf16_to_u16(reg), VEC_ELEM_NUM);
-  }
-};
-
-struct BF16Vec16 : public Vec<BF16Vec16> {
-  constexpr static int VEC_ELEM_NUM = 16;
-  fixed_vbfloat16m2_t reg;
-
-  explicit BF16Vec16(const void* ptr)
-      : reg(__riscv_vreinterpret_v_u16m2_bf16m2(__riscv_vle16_v_u16m2(
-            reinterpret_cast<const uint16_t*>(ptr), VEC_ELEM_NUM))) {};
-
-  explicit BF16Vec16(fixed_vbfloat16m2_t data) : reg(data) {};
-  explicit BF16Vec16(const FP32Vec16&);
-
-  void save(void* ptr) const {
-    __riscv_vse16_v_u16m2(reinterpret_cast<uint16_t*>(ptr), bf16_to_u16(reg),
-                          VEC_ELEM_NUM);
-  }
-  void save(void* ptr, int elem_num) const {
-    __riscv_vse16_v_u16m2(reinterpret_cast<uint16_t*>(ptr), bf16_to_u16(reg),
-                          elem_num);
-  }
-  void save_strided(void* ptr, ptrdiff_t stride) const {
-    ptrdiff_t byte_stride = stride * sizeof(uint16_t);
-    __riscv_vsse16_v_u16m2(reinterpret_cast<uint16_t*>(ptr), byte_stride,
-                           bf16_to_u16(reg), VEC_ELEM_NUM);
-  }
-};
-
-struct BF16Vec32 : public Vec<BF16Vec32> {
-  constexpr static int VEC_ELEM_NUM = 32;
-  fixed_vbfloat16m4_t reg;
-
-  explicit BF16Vec32(const void* ptr)
-      : reg(__riscv_vreinterpret_v_u16m4_bf16m4(__riscv_vle16_v_u16m4(
-            reinterpret_cast<const uint16_t*>(ptr), VEC_ELEM_NUM))) {};
-
-  explicit BF16Vec32(fixed_vbfloat16m4_t data) : reg(data) {};
-
-  explicit BF16Vec32(const BF16Vec8& v) {
-    fixed_vuint16m1_t u16_val = bf16_to_u16(v.reg);
-    fixed_vuint16m4_t u16_combined =
-        __riscv_vcreate_v_u16m1_u16m4(u16_val, u16_val, u16_val, u16_val);
-    reg = __riscv_vreinterpret_v_u16m4_bf16m4(u16_combined);
-  };
-
-  void save(void* ptr) const {
-    __riscv_vse16_v_u16m4(reinterpret_cast<uint16_t*>(ptr), bf16_to_u16(reg),
-                          VEC_ELEM_NUM);
-  }
-  void save(void* ptr, int elem_num) const {
-    __riscv_vse16_v_u16m4(reinterpret_cast<uint16_t*>(ptr), bf16_to_u16(reg),
-                          elem_num);
-  }
-  void save_strided(void* ptr, ptrdiff_t stride) const {
-    ptrdiff_t byte_stride = stride * sizeof(uint16_t);
-    __riscv_vsse16_v_u16m4(reinterpret_cast<uint16_t*>(ptr), byte_stride,
-                           bf16_to_u16(reg), VEC_ELEM_NUM);
-  }
-};
-
-#else
-// ============================================================================
-// BF16 Fallback Implementation (FP32 Simulation)
-// ============================================================================
-
-struct BF16Vec8 : public Vec<BF16Vec8> {
-  constexpr static int VEC_ELEM_NUM = 8;
-  fixed_vfloat32m2_t reg_fp32;
-  explicit BF16Vec8(const void* ptr) {
-    const uint16_t* u16 = static_cast<const uint16_t*>(ptr);
-    float tmp[8];
-    for (int i = 0; i < 8; ++i) {
-      uint32_t v = static_cast<uint32_t>(u16[i]) << 16;
-      std::memcpy(&tmp[i], &v, 4);
-    }
-    reg_fp32 = __riscv_vle32_v_f32m2(tmp, 8);
-  }
-  explicit BF16Vec8(const FP32Vec8&);
-  void save(void* ptr) const {
-    float tmp[8];
-    __riscv_vse32_v_f32m2(tmp, reg_fp32, 8);
-    uint16_t* u16 = static_cast<uint16_t*>(ptr);
-    for (int i = 0; i < 8; ++i) {
-      uint32_t v;
-      std::memcpy(&v, &tmp[i], 4);
-      u16[i] = static_cast<uint16_t>(v >> 16);
-    }
-  }
-  void save(void* ptr, int elem_num) const {
-    float tmp[8];
-    __riscv_vse32_v_f32m2(tmp, reg_fp32, 8);
-    uint16_t* u16 = static_cast<uint16_t*>(ptr);
-    for (int i = 0; i < elem_num; ++i) {
-      uint32_t v;
-      std::memcpy(&v, &tmp[i], 4);
-      u16[i] = static_cast<uint16_t>(v >> 16);
-    }
-  }
-  void save_strided(void* ptr, ptrdiff_t stride) const {
-    float tmp[8];
-    __riscv_vse32_v_f32m2(tmp, reg_fp32, 8);
-    uint8_t* u8 = static_cast<uint8_t*>(ptr);
-    ptrdiff_t byte_stride = stride * sizeof(uint16_t);
-    for (int i = 0; i < 8; ++i) {
-      uint32_t v;
-      std::memcpy(&v, &tmp[i], 4);
-      uint16_t val = static_cast<uint16_t>(v >> 16);
-      *reinterpret_cast<uint16_t*>(u8 + i * byte_stride) = val;
-    }
-  }
-};
-
-struct BF16Vec16 : public Vec<BF16Vec16> {
-  constexpr static int VEC_ELEM_NUM = 16;
-  fixed_vfloat32m4_t reg_fp32;
-  explicit BF16Vec16(const void* ptr) {
-    const uint16_t* u16 = static_cast<const uint16_t*>(ptr);
-    float tmp[16];
-    for (int i = 0; i < 16; ++i) {
-      uint32_t v = static_cast<uint32_t>(u16[i]) << 16;
-      std::memcpy(&tmp[i], &v, 4);
-    }
-    reg_fp32 = __riscv_vle32_v_f32m4(tmp, 16);
-  }
-  explicit BF16Vec16(const FP32Vec16&);
-  void save(void* ptr) const {
-    float tmp[16];
-    __riscv_vse32_v_f32m4(tmp, reg_fp32, 16);
-    uint16_t* u16 = static_cast<uint16_t*>(ptr);
-    for (int i = 0; i < 16; ++i) {
-      uint32_t v;
-      std::memcpy(&v, &tmp[i], 4);
-      u16[i] = static_cast<uint16_t>(v >> 16);
-    }
-  }
-  void save(void* ptr, int elem_num) const {
-    float tmp[16];
-    __riscv_vse32_v_f32m4(tmp, reg_fp32, 16);
-    uint16_t* u16 = static_cast<uint16_t*>(ptr);
-    for (int i = 0; i < elem_num; ++i) {
-      uint32_t v;
-      std::memcpy(&v, &tmp[i], 4);
-      u16[i] = static_cast<uint16_t>(v >> 16);
-    }
-  }
-  void save_strided(void* ptr, ptrdiff_t stride) const {
-    float tmp[16];
-    __riscv_vse32_v_f32m4(tmp, reg_fp32, 16);
-    uint8_t* u8 = static_cast<uint8_t*>(ptr);
-    ptrdiff_t byte_stride = stride * sizeof(uint16_t);
-    for (int i = 0; i < 16; ++i) {
-      uint32_t v;
-      std::memcpy(&v, &tmp[i], 4);
-      uint16_t val = static_cast<uint16_t>(v >> 16);
-      *reinterpret_cast<uint16_t*>(u8 + i * byte_stride) = val;
-    }
-  }
-};
-
-struct BF16Vec32 : public Vec<BF16Vec32> {
-  constexpr static int VEC_ELEM_NUM = 32;
-  fixed_vfloat32m8_t reg_fp32;
-
-  explicit BF16Vec32(const void* ptr) {
-    const uint16_t* u16 = static_cast<const uint16_t*>(ptr);
-    float tmp[32];
-    for (int i = 0; i < 32; ++i) {
-      uint32_t v = static_cast<uint32_t>(u16[i]) << 16;
-      std::memcpy(&tmp[i], &v, 4);
-    }
-    reg_fp32 = __riscv_vle32_v_f32m8(tmp, 32);
-  }
-
-  explicit BF16Vec32(const BF16Vec8& v) {
-    float tmp_small[8];
-    __riscv_vse32_v_f32m2(tmp_small, v.reg_fp32, 8);
-    float tmp_large[32];
-    for (int i = 0; i < 4; ++i) {
-      std::memcpy(tmp_large + (i * 8), tmp_small, 8 * sizeof(float));
-    }
-    reg_fp32 = __riscv_vle32_v_f32m8(tmp_large, 32);
-  }
-
-  void save(void* ptr) const {
-    float tmp[32];
-    __riscv_vse32_v_f32m8(tmp, reg_fp32, 32);
-    uint16_t* u16 = static_cast<uint16_t*>(ptr);
-    for (int i = 0; i < 32; ++i) {
-      uint32_t v;
-      std::memcpy(&v, &tmp[i], 4);
-      u16[i] = static_cast<uint16_t>(v >> 16);
-    }
-  }
-
-  void save(void* ptr, int elem_num) const {
-    float tmp[32];
-    __riscv_vse32_v_f32m8(tmp, reg_fp32, 32);
-    uint16_t* u16 = static_cast<uint16_t*>(ptr);
-    for (int i = 0; i < elem_num; ++i) {
-      uint32_t v;
-      std::memcpy(&v, &tmp[i], 4);
-      u16[i] = static_cast<uint16_t>(v >> 16);
-    }
-  }
-
-  void save_strided(void* ptr, ptrdiff_t stride) const {
-    float tmp[32];
-    __riscv_vse32_v_f32m8(tmp, reg_fp32, 32);
-    uint8_t* u8 = static_cast<uint8_t*>(ptr);
-    ptrdiff_t byte_stride = stride * sizeof(uint16_t);
-    for (int i = 0; i < 32; ++i) {
-      uint32_t v;
-      std::memcpy(&v, &tmp[i], 4);
-      uint16_t val = static_cast<uint16_t>(v >> 16);
-      *reinterpret_cast<uint16_t*>(u8 + i * byte_stride) = val;
-    }
-  }
-};
-#endif
-
-// ============================================================================
-// FP32 Implementation
-// ============================================================================
-
-struct FP32Vec4 : public Vec<FP32Vec4> {
-  constexpr static int VEC_ELEM_NUM = 4;
-  fixed_vfloat32m1_t reg;
-  explicit FP32Vec4(float v) : reg(__riscv_vfmv_v_f_f32m1(v, VEC_ELEM_NUM)) {};
-  explicit FP32Vec4() : reg(__riscv_vfmv_v_f_f32m1(0.0f, VEC_ELEM_NUM)) {};
-  explicit FP32Vec4(const float* ptr)
-      : reg(__riscv_vle32_v_f32m1(ptr, VEC_ELEM_NUM)) {};
-  explicit FP32Vec4(fixed_vfloat32m1_t data) : reg(data) {};
-  explicit FP32Vec4(const FP32Vec4& data) : reg(data.reg) {};
-  void save(float* ptr) const { __riscv_vse32_v_f32m1(ptr, reg, VEC_ELEM_NUM); }
-  void save(float* ptr, int elem_num) const {
-    __riscv_vse32_v_f32m1(ptr, reg, elem_num);
-  }
-};
-
-struct FP32Vec8 : public Vec<FP32Vec8> {
-  constexpr static int VEC_ELEM_NUM = 8;
-  fixed_vfloat32m2_t reg;
-
-  explicit FP32Vec8(float v) : reg(__riscv_vfmv_v_f_f32m2(v, VEC_ELEM_NUM)) {};
-  explicit FP32Vec8() : reg(__riscv_vfmv_v_f_f32m2(0.0f, VEC_ELEM_NUM)) {};
-  explicit FP32Vec8(const float* ptr)
-      : reg(__riscv_vle32_v_f32m2(ptr, VEC_ELEM_NUM)) {};
-  explicit FP32Vec8(fixed_vfloat32m2_t data) : reg(data) {};
-  explicit FP32Vec8(const FP32Vec8& data) : reg(data.reg) {};
-  explicit FP32Vec8(const FP16Vec8& v)
-      : reg(__riscv_vfwcvt_f_f_v_f32m2(v.reg, VEC_ELEM_NUM)) {};
-  explicit FP32Vec8(fixed_vfloat16m1_t v)
-      : reg(__riscv_vfwcvt_f_f_v_f32m2(v, VEC_ELEM_NUM)) {};
-
-#ifdef RISCV_BF16_SUPPORT
-  explicit FP32Vec8(fixed_vbfloat16m1_t v)
-      : reg(__riscv_vfwcvtbf16_f_f_v_f32m2(v, VEC_ELEM_NUM)) {};
-  explicit FP32Vec8(const BF16Vec8& v)
-      : reg(__riscv_vfwcvtbf16_f_f_v_f32m2(v.reg, VEC_ELEM_NUM)) {};
-#else
-  explicit FP32Vec8(const BF16Vec8& v) : reg(v.reg_fp32) {};
-#endif
-
-  float reduce_sum() const {
-    fixed_vfloat32m1_t scalar = __riscv_vfmv_s_f_f32m1(0.0f, 1);
-    scalar = __riscv_vfredusum_vs_f32m2_f32m1(reg, scalar, VEC_ELEM_NUM);
-    return __riscv_vfmv_f_s_f32m1_f32(scalar);
-  }
-
-  FP32Vec8 operator*(const FP32Vec8& b) const {
-    return FP32Vec8(__riscv_vfmul_vv_f32m2(reg, b.reg, VEC_ELEM_NUM));
-  }
-  FP32Vec8 operator+(const FP32Vec8& b) const {
-    return FP32Vec8(__riscv_vfadd_vv_f32m2(reg, b.reg, VEC_ELEM_NUM));
-  }
-  FP32Vec8 operator-(const FP32Vec8& b) const {
-    return FP32Vec8(__riscv_vfsub_vv_f32m2(reg, b.reg, VEC_ELEM_NUM));
-  }
-  FP32Vec8 operator/(const FP32Vec8& b) const {
-    return FP32Vec8(__riscv_vfdiv_vv_f32m2(reg, b.reg, VEC_ELEM_NUM));
-  }
-
-  FP32Vec8 min(const FP32Vec8& b) const {
-    return FP32Vec8(__riscv_vfmin_vv_f32m2(reg, b.reg, VEC_ELEM_NUM));
-  }
-  FP32Vec8 max(const FP32Vec8& b) const {
-    return FP32Vec8(__riscv_vfmax_vv_f32m2(reg, b.reg, VEC_ELEM_NUM));
-  }
-  FP32Vec8 abs() const {
-    return FP32Vec8(__riscv_vfabs_v_f32m2(reg, VEC_ELEM_NUM));
-  }
-
-  FP32Vec8 min(const FP32Vec8& b, int elem_num) const {
-    return FP32Vec8(__riscv_vfmin_vv_f32m2(reg, b.reg, elem_num));
-  }
-  FP32Vec8 max(const FP32Vec8& b, int elem_num) const {
-    return FP32Vec8(__riscv_vfmax_vv_f32m2(reg, b.reg, elem_num));
-  }
-
-  FP32Vec8 clamp(const FP32Vec8& min_v, const FP32Vec8& max_v) const {
-    fixed_vfloat32m2_t temp =
-        __riscv_vfmax_vv_f32m2(min_v.reg, reg, VEC_ELEM_NUM);
-    return FP32Vec8(__riscv_vfmin_vv_f32m2(max_v.reg, temp, VEC_ELEM_NUM));
-  }
-
-  void save(float* ptr) const { __riscv_vse32_v_f32m2(ptr, reg, VEC_ELEM_NUM); }
-  void save(float* ptr, int elem_num) const {
-    __riscv_vse32_v_f32m2(ptr, reg, elem_num);
-  }
-  void save_strided(float* ptr, ptrdiff_t stride) const {
-    ptrdiff_t byte_stride = stride * sizeof(float);
-    __riscv_vsse32_v_f32m2(ptr, byte_stride, reg, VEC_ELEM_NUM);
-  }
-
-  FP32Vec8 exp() const {
-    const float inv_ln2 = 1.44269504088896341f;
-    fixed_vfloat32m2_t x_scaled =
-        __riscv_vfmul_vf_f32m2(reg, inv_ln2, VEC_ELEM_NUM);
-    fixed_vint32m2_t n_int = __riscv_vfcvt_x_f_v_i32m2(x_scaled, VEC_ELEM_NUM);
-    fixed_vfloat32m2_t n_float = __riscv_vfcvt_f_x_v_f32m2(n_int, VEC_ELEM_NUM);
-
-    fixed_vfloat32m2_t r =
-        __riscv_vfsub_vv_f32m2(x_scaled, n_float, VEC_ELEM_NUM);
-
-    fixed_vfloat32m2_t poly =
-        __riscv_vfmv_v_f_f32m2(0.001333355810164f, VEC_ELEM_NUM);
-    poly = __riscv_vfmul_vv_f32m2(poly, r, VEC_ELEM_NUM);
-    poly = __riscv_vfadd_vf_f32m2(poly, 0.009618129107628f, VEC_ELEM_NUM);
-    poly = __riscv_vfmul_vv_f32m2(poly, r, VEC_ELEM_NUM);
-    poly = __riscv_vfadd_vf_f32m2(poly, 0.055504108664821f, VEC_ELEM_NUM);
-    poly = __riscv_vfmul_vv_f32m2(poly, r, VEC_ELEM_NUM);
-    poly = __riscv_vfadd_vf_f32m2(poly, 0.240226506959101f, VEC_ELEM_NUM);
-    poly = __riscv_vfmul_vv_f32m2(poly, r, VEC_ELEM_NUM);
-    poly = __riscv_vfadd_vf_f32m2(poly, 0.693147180559945f, VEC_ELEM_NUM);
-    poly = __riscv_vfmul_vv_f32m2(poly, r, VEC_ELEM_NUM);
-    poly = __riscv_vfadd_vf_f32m2(poly, 1.0f, VEC_ELEM_NUM);
-
-    fixed_vint32m2_t biased_exp =
-        __riscv_vadd_vx_i32m2(n_int, 127, VEC_ELEM_NUM);
-    biased_exp = __riscv_vmax_vx_i32m2(biased_exp, 0, VEC_ELEM_NUM);
-    fixed_vint32m2_t exponent_bits =
-        __riscv_vsll_vx_i32m2(biased_exp, 23, VEC_ELEM_NUM);
-    fixed_vfloat32m2_t scale =
-        __riscv_vreinterpret_v_i32m2_f32m2(exponent_bits);
-
-    return FP32Vec8(__riscv_vfmul_vv_f32m2(poly, scale, VEC_ELEM_NUM));
-  }
-
-  FP32Vec8 tanh() const {
-    fixed_vfloat32m2_t x_clamped = __riscv_vfmin_vf_f32m2(
-        __riscv_vfmax_vf_f32m2(reg, -9.0f, VEC_ELEM_NUM), 9.0f, VEC_ELEM_NUM);
-    fixed_vfloat32m2_t x2 =
-        __riscv_vfmul_vf_f32m2(x_clamped, 2.0f, VEC_ELEM_NUM);
-    FP32Vec8 exp_val = FP32Vec8(x2).exp();
-    fixed_vfloat32m2_t num =
-        __riscv_vfsub_vf_f32m2(exp_val.reg, 1.0f, VEC_ELEM_NUM);
-    fixed_vfloat32m2_t den =
-        __riscv_vfadd_vf_f32m2(exp_val.reg, 1.0f, VEC_ELEM_NUM);
-    return FP32Vec8(__riscv_vfdiv_vv_f32m2(num, den, VEC_ELEM_NUM));
-  }
-
-  FP32Vec8 er() const {
-    const float p = 0.3275911f, a1 = 0.254829592f, a2 = -0.284496736f,
-                a3 = 1.421413741f, a4 = -1.453152027f, a5 = 1.061405429f;
-    fixed_vfloat32m2_t abs_x = __riscv_vfabs_v_f32m2(reg, VEC_ELEM_NUM);
-
-    fixed_vfloat32m2_t t = __riscv_vfadd_vf_f32m2(
-        __riscv_vfmul_vf_f32m2(abs_x, p, VEC_ELEM_NUM), 1.0f, VEC_ELEM_NUM);
-    t = __riscv_vfrdiv_vf_f32m2(t, 1.0f, VEC_ELEM_NUM);
-
-    fixed_vfloat32m2_t poly = __riscv_vfmv_v_f_f32m2(a5, VEC_ELEM_NUM);
-    poly = __riscv_vfadd_vf_f32m2(__riscv_vfmul_vv_f32m2(poly, t, VEC_ELEM_NUM),
-                                  a4, VEC_ELEM_NUM);
-    poly = __riscv_vfadd_vf_f32m2(__riscv_vfmul_vv_f32m2(poly, t, VEC_ELEM_NUM),
-                                  a3, VEC_ELEM_NUM);
-    poly = __riscv_vfadd_vf_f32m2(__riscv_vfmul_vv_f32m2(poly, t, VEC_ELEM_NUM),
-                                  a2, VEC_ELEM_NUM);
-    poly = __riscv_vfadd_vf_f32m2(__riscv_vfmul_vv_f32m2(poly, t, VEC_ELEM_NUM),
-                                  a1, VEC_ELEM_NUM);
-    poly = __riscv_vfmul_vv_f32m2(poly, t, VEC_ELEM_NUM);
-
-    fixed_vfloat32m2_t exp_val =
-        FP32Vec8(__riscv_vfneg_v_f32m2(
-                     __riscv_vfmul_vv_f32m2(abs_x, abs_x, VEC_ELEM_NUM),
-                     VEC_ELEM_NUM))
-            .exp()
-            .reg;
-    fixed_vfloat32m2_t res = __riscv_vfrsub_vf_f32m2(
-        __riscv_vfmul_vv_f32m2(poly, exp_val, VEC_ELEM_NUM), 1.0f,
-        VEC_ELEM_NUM);
-
-    vbool16_t mask = __riscv_vmflt_vf_f32m2_b16(reg, 0.0f, VEC_ELEM_NUM);
-    return FP32Vec8(__riscv_vfneg_v_f32m2_m(mask, res, VEC_ELEM_NUM));
-  }
-};
-
-struct FP32Vec16 : public Vec<FP32Vec16> {
-  constexpr static int VEC_ELEM_NUM = 16;
-  fixed_vfloat32m4_t reg;
-
-  explicit FP32Vec16(float v) : reg(__riscv_vfmv_v_f_f32m4(v, VEC_ELEM_NUM)) {};
-  explicit FP32Vec16() : reg(__riscv_vfmv_v_f_f32m4(0.0f, VEC_ELEM_NUM)) {};
-  explicit FP32Vec16(const float* ptr)
-      : reg(__riscv_vle32_v_f32m4(ptr, VEC_ELEM_NUM)) {};
-  explicit FP32Vec16(fixed_vfloat32m4_t data) : reg(data) {};
-  explicit FP32Vec16(const FP32Vec8& data)
-      : reg(__riscv_vcreate_v_f32m2_f32m4(data.reg, data.reg)) {};
-  explicit FP32Vec16(const FP32Vec16& data) : reg(data.reg) {};
-  explicit FP32Vec16(const FP16Vec16& v);
-
-#ifdef RISCV_BF16_SUPPORT
-  explicit FP32Vec16(fixed_vbfloat16m2_t v)
-      : reg(__riscv_vfwcvtbf16_f_f_v_f32m4(v, VEC_ELEM_NUM)) {};
-  explicit FP32Vec16(const BF16Vec16& v)
-      : reg(__riscv_vfwcvtbf16_f_f_v_f32m4(v.reg, VEC_ELEM_NUM)) {};
-#else
-  explicit FP32Vec16(const BF16Vec16& v) : reg(v.reg_fp32) {};
-#endif
-
-  FP32Vec16 operator+(const FP32Vec16& b) const {
-    return FP32Vec16(__riscv_vfadd_vv_f32m4(reg, b.reg, VEC_ELEM_NUM));
-  }
-  FP32Vec16 operator-(const FP32Vec16& b) const {
-    return FP32Vec16(__riscv_vfsub_vv_f32m4(reg, b.reg, VEC_ELEM_NUM));
-  }
-  FP32Vec16 operator*(const FP32Vec16& b) const {
-    return FP32Vec16(__riscv_vfmul_vv_f32m4(reg, b.reg, VEC_ELEM_NUM));
-  }
-  FP32Vec16 operator/(const FP32Vec16& b) const {
-    return FP32Vec16(__riscv_vfdiv_vv_f32m4(reg, b.reg, VEC_ELEM_NUM));
-  }
-
-  FP32Vec16 fma(const FP32Vec16& a, const FP32Vec16& b) const {
-    return FP32Vec16(__riscv_vfmacc_vv_f32m4(reg, a.reg, b.reg, VEC_ELEM_NUM));
-  }
-
-  float reduce_sum() const {
-    fixed_vfloat32m1_t scalar = __riscv_vfmv_s_f_f32m1(0.0f, 1);
-    scalar = __riscv_vfredusum_vs_f32m4_f32m1(reg, scalar, VEC_ELEM_NUM);
-    return __riscv_vfmv_f_s_f32m1_f32(scalar);
-  }
-
-  float reduce_max() const {
-    fixed_vfloat32m1_t scalar =
-        __riscv_vfmv_s_f_f32m1(std::numeric_limits<float>::lowest(), 1);
-    scalar = __riscv_vfredmax_vs_f32m4_f32m1(reg, scalar, VEC_ELEM_NUM);
-    return __riscv_vfmv_f_s_f32m1_f32(scalar);
-  }
-
-  float reduce_min() const {
-    fixed_vfloat32m1_t scalar =
-        __riscv_vfmv_s_f_f32m1(std::numeric_limits<float>::max(), 1);
-    scalar = __riscv_vfredmin_vs_f32m4_f32m1(reg, scalar, VEC_ELEM_NUM);
-    return __riscv_vfmv_f_s_f32m1_f32(scalar);
-  }
-
-  template <int group_size>
-  float reduce_sub_sum(int idx) {
-    static_assert(VEC_ELEM_NUM % group_size == 0);
-    const int start = idx * group_size;
-    vuint32m4_t indices = __riscv_vid_v_u32m4(VEC_ELEM_NUM);
-    vbool8_t mask = __riscv_vmand_mm_b8(
-        __riscv_vmsgeu_vx_u32m4_b8(indices, start, VEC_ELEM_NUM),
-        __riscv_vmsltu_vx_u32m4_b8(indices, start + group_size, VEC_ELEM_NUM),
-        VEC_ELEM_NUM);
-    fixed_vfloat32m1_t scalar = __riscv_vfmv_s_f_f32m1(0.0f, 1);
-    scalar =
-        __riscv_vfredusum_vs_f32m4_f32m1_m(mask, reg, scalar, VEC_ELEM_NUM);
-    return __riscv_vfmv_f_s_f32m1_f32(scalar);
-  };
-
-  FP32Vec16 max(const FP32Vec16& b) const {
-    return FP32Vec16(__riscv_vfmax_vv_f32m4(reg, b.reg, VEC_ELEM_NUM));
-  }
-  FP32Vec16 min(const FP32Vec16& b) const {
-    return FP32Vec16(__riscv_vfmin_vv_f32m4(reg, b.reg, VEC_ELEM_NUM));
-  }
-  FP32Vec16 abs() const {
-    return FP32Vec16(__riscv_vfabs_v_f32m4(reg, VEC_ELEM_NUM));
-  }
-
-  FP32Vec16 clamp(const FP32Vec16& min_v, const FP32Vec16& max_v) const {
-    return FP32Vec16(__riscv_vfmin_vv_f32m4(
-        max_v.reg, __riscv_vfmax_vv_f32m4(min_v.reg, reg, VEC_ELEM_NUM),
-        VEC_ELEM_NUM));
-  }
-
-  void save(float* ptr) const { __riscv_vse32_v_f32m4(ptr, reg, VEC_ELEM_NUM); }
-  void save(float* ptr, int elem_num) const {
-    __riscv_vse32_v_f32m4(ptr, reg, elem_num);
-  }
-  void save_strided(float* ptr, ptrdiff_t stride) const {
-    ptrdiff_t byte_stride = stride * sizeof(float);
-    __riscv_vsse32_v_f32m4(ptr, byte_stride, reg, VEC_ELEM_NUM);
-  }
-
-  FP32Vec16 exp() const {
-    const float inv_ln2 = 1.44269504088896341f;
-    fixed_vfloat32m4_t x_scaled =
-        __riscv_vfmul_vf_f32m4(reg, inv_ln2, VEC_ELEM_NUM);
-    fixed_vint32m4_t n_int = __riscv_vfcvt_x_f_v_i32m4(x_scaled, VEC_ELEM_NUM);
-    fixed_vfloat32m4_t n_float = __riscv_vfcvt_f_x_v_f32m4(n_int, VEC_ELEM_NUM);
-    fixed_vfloat32m4_t r =
-        __riscv_vfsub_vv_f32m4(x_scaled, n_float, VEC_ELEM_NUM);
-
-    fixed_vfloat32m4_t poly =
-        __riscv_vfmv_v_f_f32m4(0.001333355810164f, VEC_ELEM_NUM);
-    poly = __riscv_vfadd_vf_f32m4(__riscv_vfmul_vv_f32m4(poly, r, VEC_ELEM_NUM),
-                                  0.009618129107628f, VEC_ELEM_NUM);
-    poly = __riscv_vfadd_vf_f32m4(__riscv_vfmul_vv_f32m4(poly, r, VEC_ELEM_NUM),
-                                  0.055504108664821f, VEC_ELEM_NUM);
-    poly = __riscv_vfadd_vf_f32m4(__riscv_vfmul_vv_f32m4(poly, r, VEC_ELEM_NUM),
-                                  0.240226506959101f, VEC_ELEM_NUM);
-    poly = __riscv_vfadd_vf_f32m4(__riscv_vfmul_vv_f32m4(poly, r, VEC_ELEM_NUM),
-                                  0.693147180559945f, VEC_ELEM_NUM);
-    poly = __riscv_vfadd_vf_f32m4(__riscv_vfmul_vv_f32m4(poly, r, VEC_ELEM_NUM),
-                                  1.0f, VEC_ELEM_NUM);
-
-    fixed_vint32m4_t biased_exp = __riscv_vmax_vx_i32m4(
-        __riscv_vadd_vx_i32m4(n_int, 127, VEC_ELEM_NUM), 0, VEC_ELEM_NUM);
-    fixed_vfloat32m4_t scale = __riscv_vreinterpret_v_i32m4_f32m4(
-        __riscv_vsll_vx_i32m4(biased_exp, 23, VEC_ELEM_NUM));
-
-    return FP32Vec16(__riscv_vfmul_vv_f32m4(poly, scale, VEC_ELEM_NUM));
-  }
-
-  FP32Vec16 tanh() const {
-    fixed_vfloat32m4_t x_clamped = __riscv_vfmin_vf_f32m4(
-        __riscv_vfmax_vf_f32m4(reg, -9.0f, VEC_ELEM_NUM), 9.0f, VEC_ELEM_NUM);
-    FP32Vec16 exp_val =
-        FP32Vec16(__riscv_vfmul_vf_f32m4(x_clamped, 2.0f, VEC_ELEM_NUM)).exp();
-    return FP32Vec16(__riscv_vfdiv_vv_f32m4(
-        __riscv_vfsub_vf_f32m4(exp_val.reg, 1.0f, VEC_ELEM_NUM),
-        __riscv_vfadd_vf_f32m4(exp_val.reg, 1.0f, VEC_ELEM_NUM), VEC_ELEM_NUM));
-  }
-
-  FP32Vec16 er() const {
-    const float p = 0.3275911f, a1 = 0.254829592f, a2 = -0.284496736f,
-                a3 = 1.421413741f, a4 = -1.453152027f, a5 = 1.061405429f;
-    fixed_vfloat32m4_t abs_x = __riscv_vfabs_v_f32m4(reg, VEC_ELEM_NUM);
-    fixed_vfloat32m4_t t = __riscv_vfrdiv_vf_f32m4(
-        __riscv_vfadd_vf_f32m4(__riscv_vfmul_vf_f32m4(abs_x, p, VEC_ELEM_NUM),
-                               1.0f, VEC_ELEM_NUM),
-        1.0f, VEC_ELEM_NUM);
-
-    fixed_vfloat32m4_t poly = __riscv_vfmv_v_f_f32m4(a5, VEC_ELEM_NUM);
-    poly = __riscv_vfadd_vf_f32m4(__riscv_vfmul_vv_f32m4(poly, t, VEC_ELEM_NUM),
-                                  a4, VEC_ELEM_NUM);
-    poly = __riscv_vfadd_vf_f32m4(__riscv_vfmul_vv_f32m4(poly, t, VEC_ELEM_NUM),
-                                  a3, VEC_ELEM_NUM);
-    poly = __riscv_vfadd_vf_f32m4(__riscv_vfmul_vv_f32m4(poly, t, VEC_ELEM_NUM),
-                                  a2, VEC_ELEM_NUM);
-    poly = __riscv_vfadd_vf_f32m4(__riscv_vfmul_vv_f32m4(poly, t, VEC_ELEM_NUM),
-                                  a1, VEC_ELEM_NUM);
-    poly = __riscv_vfmul_vv_f32m4(poly, t, VEC_ELEM_NUM);
-
-    fixed_vfloat32m4_t exp_val =
-        FP32Vec16(__riscv_vfneg_v_f32m4(
-                      __riscv_vfmul_vv_f32m4(abs_x, abs_x, VEC_ELEM_NUM),
-                      VEC_ELEM_NUM))
-            .exp()
-            .reg;
-    fixed_vfloat32m4_t res = __riscv_vfrsub_vf_f32m4(
-        __riscv_vfmul_vv_f32m4(poly, exp_val, VEC_ELEM_NUM), 1.0f,
-        VEC_ELEM_NUM);
-
-    vbool8_t mask = __riscv_vmflt_vf_f32m4_b8(reg, 0.0f, VEC_ELEM_NUM);
-    return FP32Vec16(__riscv_vfneg_v_f32m4_m(mask, res, VEC_ELEM_NUM));
-  }
-};
-
-// ============================================================================
-// Type Traits & Global Helpers
-// ============================================================================
-
-template <typename T>
-struct VecType {
-  using vec_type = void;
-  using vec_t = void;
-};
-
-template <typename T>
-using vec_t = typename VecType<T>::vec_type;
-
-template <>
-struct VecType<float> {
-  using vec_type = FP32Vec8;
-  using vec_t = FP32Vec8;
-};
-template <>
-struct VecType<c10::Half> {
-  using vec_type = FP16Vec8;
-  using vec_t = FP16Vec8;
-};
-template <>
-struct VecType<c10::BFloat16> {
-  using vec_type = BF16Vec8;
-  using vec_t = BF16Vec8;
-};
-
-template <typename T>
-void storeFP32(float v, T* ptr) {
-  *ptr = v;
-}
-template <>
-inline void storeFP32<c10::Half>(float v, c10::Half* ptr) {
-  *reinterpret_cast<_Float16*>(ptr) = static_cast<_Float16>(v);
-}
-
-inline FP16Vec16::FP16Vec16(const FP32Vec16& v) {
-  reg = __riscv_vfncvt_f_f_w_f16m2(v.reg, VEC_ELEM_NUM);
-}
-inline FP16Vec8::FP16Vec8(const FP32Vec8& v) {
-  reg = __riscv_vfncvt_f_f_w_f16m1(v.reg, VEC_ELEM_NUM);
-}
-inline FP32Vec16::FP32Vec16(const FP16Vec16& v) {
-  reg = __riscv_vfwcvt_f_f_v_f32m4(v.reg, VEC_ELEM_NUM);
-}
-inline void fma(FP32Vec16& acc, const FP32Vec16& a, const FP32Vec16& b) {
-  acc = acc.fma(a, b);
-}
-
-#ifdef RISCV_BF16_SUPPORT
-template <>
-inline void storeFP32<c10::BFloat16>(float v, c10::BFloat16* ptr) {
-  *ptr = static_cast<__bf16>(v);
-};
-inline BF16Vec8::BF16Vec8(const FP32Vec8& v)
-    : reg(__riscv_vfncvtbf16_f_f_w_bf16m1(v.reg, VEC_ELEM_NUM)) {};
-inline BF16Vec16::BF16Vec16(const FP32Vec16& v)
-    : reg(__riscv_vfncvtbf16_f_f_w_bf16m2(v.reg, VEC_ELEM_NUM)) {};
-#else
-template <>
-inline void storeFP32<c10::BFloat16>(float v, c10::BFloat16* ptr) {
-  uint32_t val;
-  std::memcpy(&val, &v, 4);
-  *reinterpret_cast<uint16_t*>(ptr) = static_cast<uint16_t>(val >> 16);
-}
-inline BF16Vec8::BF16Vec8(const FP32Vec8& v) : reg_fp32(v.reg) {}
-inline BF16Vec16::BF16Vec16(const FP32Vec16& v) : reg_fp32(v.reg) {}
-#endif
-
-inline void prefetch(const void* addr) { __builtin_prefetch(addr, 0, 1); }
-
-}  // namespace vec_op
-
-#ifndef CPU_KERNEL_GUARD_IN
-  #define CPU_KERNEL_GUARD_IN(NAME)
-#endif
-
-#ifndef CPU_KERNEL_GUARD_OUT
-  #define CPU_KERNEL_GUARD_OUT(NAME)
-#endif
-
-#endif  // CPU_TYPES_RISCV_HPP
+#endif  // CPU_TYPES_RISCV_HPP

csrc/cpu/cpu_types_riscv_defs.hpp

@@ -0,0 +1,98 @@
+#ifndef CPU_TYPES_RISCV_DEFS_HPP
+#define CPU_TYPES_RISCV_DEFS_HPP
+
+// VLEN-to-LMUL mapping for RISC-V Vector extension.
+//
+// LMUL_<N> expands to the LMUL suffix giving N total bits of vector data:
+//   VLEN=128: LMUL_128=m1,  LMUL_256=m2,  LMUL_512=m4,  LMUL_1024=m8
+//   VLEN=256: LMUL_128=mf2, LMUL_256=m1,  LMUL_512=m2,  LMUL_1024=m4
+
+#include <riscv_vector.h>
+
+#if __riscv_v_min_vlen == 128
+  #define LMUL_128 m1
+  #define LMUL_256 m2
+  #define LMUL_512 m4
+  #define LMUL_1024 m8
+  #define BOOL_256 b16
+  #define BOOL_512 b8
+#elif __riscv_v_min_vlen == 256
+  #define LMUL_128 mf2
+  #define LMUL_256 m1
+  #define LMUL_512 m2
+  #define LMUL_1024 m4
+  #define BOOL_256 b32
+  #define BOOL_512 b16
+#else
+  #error "cpu_types_riscv_defs.hpp: unsupported __riscv_v_min_vlen"
+#endif
+
+// Token-paste helpers.
+#define _RVV_P2(a, b) a##b
+#define _RVV_P3(a, b, c) a##b##c
+#define _RVV_P4(a, b, c, d) a##b##c##d
+#define RVVTYPE(base, lmul, suffix) _RVV_P3(base, lmul, suffix)
+#define RVVI(base, lmul) _RVV_P2(base, lmul)
+#define RVVI3(base, lmul, suffix) _RVV_P3(base, lmul, suffix)
+#define RVVI4(a, b, c, d) _RVV_P4(a, b, c, d)
+// For mask intrinsics: RVVIB(base, LMUL_256, BOOL_256) → base##m2##_##b16
+#define _RVV_PB(base, lmul, btype) base##lmul##_##btype
+#define RVVIB(base, lmul, btype) _RVV_PB(base, lmul, btype)
+
+// ---- Semantic fixed-vector typedefs (named by element count) ----
+
+// float16
+typedef RVVTYPE(vfloat16, LMUL_128, _t) fixed_fp16x8_t
+    __attribute__((riscv_rvv_vector_bits(128)));
+typedef RVVTYPE(vfloat16, LMUL_256, _t) fixed_fp16x16_t
+    __attribute__((riscv_rvv_vector_bits(256)));
+
+// float32
+typedef RVVTYPE(vfloat32, LMUL_128, _t) fixed_fp32x4_t
+    __attribute__((riscv_rvv_vector_bits(128)));
+typedef RVVTYPE(vfloat32, LMUL_256, _t) fixed_fp32x8_t
+    __attribute__((riscv_rvv_vector_bits(256)));
+typedef RVVTYPE(vfloat32, LMUL_512, _t) fixed_fp32x16_t
+    __attribute__((riscv_rvv_vector_bits(512)));
+typedef RVVTYPE(vfloat32, LMUL_1024, _t) fixed_fp32x32_t
+    __attribute__((riscv_rvv_vector_bits(1024)));
+
+// int32
+typedef RVVTYPE(vint32, LMUL_256, _t) fixed_i32x8_t
+    __attribute__((riscv_rvv_vector_bits(256)));
+typedef RVVTYPE(vint32, LMUL_512, _t) fixed_i32x16_t
+    __attribute__((riscv_rvv_vector_bits(512)));
+
+// uint16
+typedef RVVTYPE(vuint16, LMUL_128, _t) fixed_u16x8_t
+    __attribute__((riscv_rvv_vector_bits(128)));
+typedef RVVTYPE(vuint16, LMUL_256, _t) fixed_u16x16_t
+    __attribute__((riscv_rvv_vector_bits(256)));
+typedef RVVTYPE(vuint16, LMUL_512, _t) fixed_u16x32_t
+    __attribute__((riscv_rvv_vector_bits(512)));
+
+// bfloat16
+#ifdef RISCV_BF16_SUPPORT
+typedef RVVTYPE(vbfloat16, LMUL_128, _t) fixed_bf16x8_t
+    __attribute__((riscv_rvv_vector_bits(128)));
+typedef RVVTYPE(vbfloat16, LMUL_256, _t) fixed_bf16x16_t
+    __attribute__((riscv_rvv_vector_bits(256)));
+typedef RVVTYPE(vbfloat16, LMUL_512, _t) fixed_bf16x32_t
+    __attribute__((riscv_rvv_vector_bits(512)));
+#endif
+
+// ---- Reduction accumulator type (always m1 = one register of f32) ----
+// Used for scalar reductions; only element [0] is meaningful.
+typedef vfloat32m1_t rvv_f32_accum_t
+    __attribute__((riscv_rvv_vector_bits(__riscv_v_min_vlen)));
+
+// ---- Mask types for f32 elements ----
+#if __riscv_v_min_vlen == 128
+typedef vbool16_t rvv_mask_f32x8_t;
+typedef vbool8_t rvv_mask_f32x16_t;
+#elif __riscv_v_min_vlen == 256
+typedef vbool32_t rvv_mask_f32x8_t;
+typedef vbool16_t rvv_mask_f32x16_t;
+#endif
+
+#endif  // CPU_TYPES_RISCV_DEFS_HPP

csrc/cpu/cpu_types_riscv_impl.hpp

@@ -0,0 +1,905 @@
+#ifndef CPU_TYPES_RISCV_IMPL_HPP
+#define CPU_TYPES_RISCV_IMPL_HPP
+
+// Shared implementation of RVV vector-type wrapper classes.
+// This file is VLEN-independent: it uses the semantic type names and
+// RVVI() intrinsic macros from cpu_types_riscv_defs.hpp.
+//
+// DO NOT include this file directly; include cpu_types_riscv.hpp instead.
+
+#include <algorithm>
+#include <cmath>
+#include <cstring>
+#include <iostream>
+#include <limits>
+#include <torch/all.h>
+namespace vec_op {
+
+// BFloat16 is always supported on RISC-V: natively when RISCV_BF16_SUPPORT
+// is defined, otherwise via the FP32-simulation fallback path.
+#define VLLM_DISPATCH_CASE_FLOATING_TYPES(...)         \
+  AT_DISPATCH_CASE(at::ScalarType::Float, __VA_ARGS__) \
+  AT_DISPATCH_CASE(at::ScalarType::Half, __VA_ARGS__)  \
+  AT_DISPATCH_CASE(at::ScalarType::BFloat16, __VA_ARGS__)
+
+#define VLLM_DISPATCH_FLOATING_TYPES(TYPE, NAME, ...) \
+  AT_DISPATCH_SWITCH(TYPE, NAME, VLLM_DISPATCH_CASE_FLOATING_TYPES(__VA_ARGS__))
+
+#define FORCE_INLINE __attribute__((always_inline)) inline
+
+namespace {
+template <typename T, T... indexes, typename F>
+constexpr void unroll_loop_item(std::integer_sequence<T, indexes...>, F&& f) {
+  (f(std::integral_constant<T, indexes>{}), ...);
+};
+}  // namespace
+
+template <typename T, T count, typename F,
+          typename = std::enable_if_t<std::is_invocable_v<F, T>>>
+constexpr void unroll_loop(F&& f) {
+  unroll_loop_item(std::make_integer_sequence<T, count>{}, std::forward<F>(f));
+}
+
+template <typename T>
+struct Vec {
+  constexpr static int get_elem_num() { return T::VEC_ELEM_NUM; };
+};
+
+struct FP32Vec8;
+struct FP32Vec16;
+
+// ============================================================================
+// FP16 Implementation
+// ============================================================================
+
+struct FP16Vec8 : public Vec<FP16Vec8> {
+  constexpr static int VEC_ELEM_NUM = 8;
+  fixed_fp16x8_t reg;
+
+  explicit FP16Vec8(const void* ptr)
+      : reg(RVVI(__riscv_vle16_v_f16, LMUL_128)(
+            static_cast<const _Float16*>(ptr), VEC_ELEM_NUM)) {};
+
+  explicit FP16Vec8(const FP32Vec8&);
+
+  void save(void* ptr) const {
+    RVVI(__riscv_vse16_v_f16, LMUL_128)(static_cast<_Float16*>(ptr), reg,
+                                        VEC_ELEM_NUM);
+  }
+  void save(void* ptr, int elem_num) const {
+    RVVI(__riscv_vse16_v_f16, LMUL_128)(static_cast<_Float16*>(ptr), reg,
+                                        elem_num);
+  }
+  void save_strided(void* ptr, ptrdiff_t stride) const {
+    ptrdiff_t byte_stride = stride * sizeof(_Float16);
+    RVVI(__riscv_vsse16_v_f16, LMUL_128)(static_cast<_Float16*>(ptr),
+                                         byte_stride, reg, VEC_ELEM_NUM);
+  }
+};
+
+struct FP16Vec16 : public Vec<FP16Vec16> {
+  constexpr static int VEC_ELEM_NUM = 16;
+  fixed_fp16x16_t reg;
+
+  explicit FP16Vec16(const void* ptr)
+      : reg(RVVI(__riscv_vle16_v_f16, LMUL_256)(
+            static_cast<const _Float16*>(ptr), VEC_ELEM_NUM)) {};
+
+  explicit FP16Vec16(const FP32Vec16& vec);
+
+  void save(void* ptr) const {
+    RVVI(__riscv_vse16_v_f16, LMUL_256)(static_cast<_Float16*>(ptr), reg,
+                                        VEC_ELEM_NUM);
+  }
+  void save(void* ptr, int elem_num) const {
+    RVVI(__riscv_vse16_v_f16, LMUL_256)(static_cast<_Float16*>(ptr), reg,
+                                        elem_num);
+  }
+  void save_strided(void* ptr, ptrdiff_t stride) const {
+    ptrdiff_t byte_stride = stride * sizeof(_Float16);
+    RVVI(__riscv_vsse16_v_f16, LMUL_256)(static_cast<_Float16*>(ptr),
+                                         byte_stride, reg, VEC_ELEM_NUM);
+  }
+};
+
+// ============================================================================
+// BF16 Implementation
+// ============================================================================
+
+#ifdef RISCV_BF16_SUPPORT
+
+FORCE_INLINE fixed_u16x8_t bf16_to_u16(fixed_bf16x8_t v) {
+  return RVVI4(__riscv_vreinterpret_v_bf16, LMUL_128, _u16, LMUL_128)(v);
+}
+FORCE_INLINE fixed_u16x16_t bf16_to_u16(fixed_bf16x16_t v) {
+  return RVVI4(__riscv_vreinterpret_v_bf16, LMUL_256, _u16, LMUL_256)(v);
+}
+FORCE_INLINE fixed_u16x32_t bf16_to_u16(fixed_bf16x32_t v) {
+  return RVVI4(__riscv_vreinterpret_v_bf16, LMUL_512, _u16, LMUL_512)(v);
+}
+
+struct BF16Vec8 : public Vec<BF16Vec8> {
+  constexpr static int VEC_ELEM_NUM = 8;
+  fixed_bf16x8_t reg;
+
+  explicit BF16Vec8(const void* ptr)
+      : reg(RVVI4(__riscv_vreinterpret_v_u16, LMUL_128, _bf16,
+                  LMUL_128)(RVVI(__riscv_vle16_v_u16, LMUL_128)(
+            reinterpret_cast<const uint16_t*>(ptr), VEC_ELEM_NUM))) {};
+
+  explicit BF16Vec8(fixed_bf16x8_t data) : reg(data) {};
+  explicit BF16Vec8(const FP32Vec8&);
+
+  void save(void* ptr) const {
+    RVVI(__riscv_vse16_v_u16, LMUL_128)(reinterpret_cast<uint16_t*>(ptr),
+                                        bf16_to_u16(reg), VEC_ELEM_NUM);
+  }
+  void save(void* ptr, int elem_num) const {
+    RVVI(__riscv_vse16_v_u16, LMUL_128)(reinterpret_cast<uint16_t*>(ptr),
+                                        bf16_to_u16(reg), elem_num);
+  }
+  void save_strided(void* ptr, ptrdiff_t stride) const {
+    ptrdiff_t byte_stride = stride * sizeof(uint16_t);
+    RVVI(__riscv_vsse16_v_u16, LMUL_128)(reinterpret_cast<uint16_t*>(ptr),
+                                         byte_stride, bf16_to_u16(reg),
+                                         VEC_ELEM_NUM);
+  }
+};
+
+struct BF16Vec16 : public Vec<BF16Vec16> {
+  constexpr static int VEC_ELEM_NUM = 16;
+  fixed_bf16x16_t reg;
+
+  explicit BF16Vec16(const void* ptr)
+      : reg(RVVI4(__riscv_vreinterpret_v_u16, LMUL_256, _bf16,
+                  LMUL_256)(RVVI(__riscv_vle16_v_u16, LMUL_256)(
+            reinterpret_cast<const uint16_t*>(ptr), VEC_ELEM_NUM))) {};
+
+  explicit BF16Vec16(fixed_bf16x16_t data) : reg(data) {};
+  explicit BF16Vec16(const FP32Vec16&);
+
+  void save(void* ptr) const {
+    RVVI(__riscv_vse16_v_u16, LMUL_256)(reinterpret_cast<uint16_t*>(ptr),
+                                        bf16_to_u16(reg), VEC_ELEM_NUM);
+  }
+  void save(void* ptr, int elem_num) const {
+    RVVI(__riscv_vse16_v_u16, LMUL_256)(reinterpret_cast<uint16_t*>(ptr),
+                                        bf16_to_u16(reg), elem_num);
+  }
+  void save_strided(void* ptr, ptrdiff_t stride) const {
+    ptrdiff_t byte_stride = stride * sizeof(uint16_t);
+    RVVI(__riscv_vsse16_v_u16, LMUL_256)(reinterpret_cast<uint16_t*>(ptr),
+                                         byte_stride, bf16_to_u16(reg),
+                                         VEC_ELEM_NUM);
+  }
+};
+
+struct BF16Vec32 : public Vec<BF16Vec32> {
+  constexpr static int VEC_ELEM_NUM = 32;
+  fixed_bf16x32_t reg;
+
+  explicit BF16Vec32(const void* ptr)
+      : reg(RVVI4(__riscv_vreinterpret_v_u16, LMUL_512, _bf16,
+                  LMUL_512)(RVVI(__riscv_vle16_v_u16, LMUL_512)(
+            reinterpret_cast<const uint16_t*>(ptr), VEC_ELEM_NUM))) {};
+
+  explicit BF16Vec32(fixed_bf16x32_t data) : reg(data) {};
+
+  explicit BF16Vec32(const BF16Vec8& v) {
+    fixed_u16x8_t u16_val = bf16_to_u16(v.reg);
+    fixed_u16x32_t u16_combined =
+        RVVI4(__riscv_vcreate_v_u16, LMUL_128, _u16, LMUL_512)(
+            u16_val, u16_val, u16_val, u16_val);
+    reg = RVVI4(__riscv_vreinterpret_v_u16, LMUL_512, _bf16,
+                LMUL_512)(u16_combined);
+  };
+
+  void save(void* ptr) const {
+    RVVI(__riscv_vse16_v_u16, LMUL_512)(reinterpret_cast<uint16_t*>(ptr),
+                                        bf16_to_u16(reg), VEC_ELEM_NUM);
+  }
+  void save(void* ptr, int elem_num) const {
+    RVVI(__riscv_vse16_v_u16, LMUL_512)(reinterpret_cast<uint16_t*>(ptr),
+                                        bf16_to_u16(reg), elem_num);
+  }
+  void save_strided(void* ptr, ptrdiff_t stride) const {
+    ptrdiff_t byte_stride = stride * sizeof(uint16_t);
+    RVVI(__riscv_vsse16_v_u16, LMUL_512)(reinterpret_cast<uint16_t*>(ptr),
+                                         byte_stride, bf16_to_u16(reg),
+                                         VEC_ELEM_NUM);
+  }
+};
+
+#else
+// ============================================================================
+// BF16 Fallback Implementation (FP32 Simulation)
+// ============================================================================
+
+struct BF16Vec8 : public Vec<BF16Vec8> {
+  constexpr static int VEC_ELEM_NUM = 8;
+  fixed_fp32x8_t reg_fp32;
+  explicit BF16Vec8(const void* ptr) {
+    const uint16_t* u16 = static_cast<const uint16_t*>(ptr);
+    float tmp[8];
+    for (int i = 0; i < 8; ++i) {
+      uint32_t v = static_cast<uint32_t>(u16[i]) << 16;
+      std::memcpy(&tmp[i], &v, 4);
+    }
+    reg_fp32 = RVVI(__riscv_vle32_v_f32, LMUL_256)(tmp, 8);
+  }
+  explicit BF16Vec8(const FP32Vec8&);
+  void save(void* ptr) const {
+    float tmp[8];
+    RVVI(__riscv_vse32_v_f32, LMUL_256)(tmp, reg_fp32, 8);
+    uint16_t* u16 = static_cast<uint16_t*>(ptr);
+    for (int i = 0; i < 8; ++i) {
+      uint32_t v;
+      std::memcpy(&v, &tmp[i], 4);
+      u16[i] = static_cast<uint16_t>(v >> 16);
+    }
+  }
+  void save(void* ptr, int elem_num) const {
+    float tmp[8];
+    RVVI(__riscv_vse32_v_f32, LMUL_256)(tmp, reg_fp32, 8);
+    uint16_t* u16 = static_cast<uint16_t*>(ptr);
+    for (int i = 0; i < elem_num; ++i) {
+      uint32_t v;
+      std::memcpy(&v, &tmp[i], 4);
+      u16[i] = static_cast<uint16_t>(v >> 16);
+    }
+  }
+  void save_strided(void* ptr, ptrdiff_t stride) const {
+    float tmp[8];
+    RVVI(__riscv_vse32_v_f32, LMUL_256)(tmp, reg_fp32, 8);
+    uint8_t* u8 = static_cast<uint8_t*>(ptr);
+    ptrdiff_t byte_stride = stride * sizeof(uint16_t);
+    for (int i = 0; i < 8; ++i) {
+      uint32_t v;
+      std::memcpy(&v, &tmp[i], 4);
+      uint16_t val = static_cast<uint16_t>(v >> 16);
+      *reinterpret_cast<uint16_t*>(u8 + i * byte_stride) = val;
+    }
+  }
+};
+
+struct BF16Vec16 : public Vec<BF16Vec16> {
+  constexpr static int VEC_ELEM_NUM = 16;
+  fixed_fp32x16_t reg_fp32;
+  explicit BF16Vec16(const void* ptr) {
+    const uint16_t* u16 = static_cast<const uint16_t*>(ptr);
+    float tmp[16];
+    for (int i = 0; i < 16; ++i) {
+      uint32_t v = static_cast<uint32_t>(u16[i]) << 16;
+      std::memcpy(&tmp[i], &v, 4);
+    }
+    reg_fp32 = RVVI(__riscv_vle32_v_f32, LMUL_512)(tmp, 16);
+  }
+  explicit BF16Vec16(const FP32Vec16&);
+  void save(void* ptr) const {
+    float tmp[16];
+    RVVI(__riscv_vse32_v_f32, LMUL_512)(tmp, reg_fp32, 16);
+    uint16_t* u16 = static_cast<uint16_t*>(ptr);
+    for (int i = 0; i < 16; ++i) {
+      uint32_t v;
+      std::memcpy(&v, &tmp[i], 4);
+      u16[i] = static_cast<uint16_t>(v >> 16);
+    }
+  }
+  void save(void* ptr, int elem_num) const {
+    float tmp[16];
+    RVVI(__riscv_vse32_v_f32, LMUL_512)(tmp, reg_fp32, 16);
+    uint16_t* u16 = static_cast<uint16_t*>(ptr);
+    for (int i = 0; i < elem_num; ++i) {
+      uint32_t v;
+      std::memcpy(&v, &tmp[i], 4);
+      u16[i] = static_cast<uint16_t>(v >> 16);
+    }
+  }
+  void save_strided(void* ptr, ptrdiff_t stride) const {
+    float tmp[16];
+    RVVI(__riscv_vse32_v_f32, LMUL_512)(tmp, reg_fp32, 16);
+    uint8_t* u8 = static_cast<uint8_t*>(ptr);
+    ptrdiff_t byte_stride = stride * sizeof(uint16_t);
+    for (int i = 0; i < 16; ++i) {
+      uint32_t v;
+      std::memcpy(&v, &tmp[i], 4);
+      uint16_t val = static_cast<uint16_t>(v >> 16);
+      *reinterpret_cast<uint16_t*>(u8 + i * byte_stride) = val;
+    }
+  }
+};
+
+struct BF16Vec32 : public Vec<BF16Vec32> {
+  constexpr static int VEC_ELEM_NUM = 32;
+  fixed_fp32x32_t reg_fp32;
+
+  explicit BF16Vec32(const void* ptr) {
+    const uint16_t* u16 = static_cast<const uint16_t*>(ptr);
+    float tmp[32];
+    for (int i = 0; i < 32; ++i) {
+      uint32_t v = static_cast<uint32_t>(u16[i]) << 16;
+      std::memcpy(&tmp[i], &v, 4);
+    }
+    reg_fp32 = RVVI(__riscv_vle32_v_f32, LMUL_1024)(tmp, 32);
+  }
+
+  explicit BF16Vec32(const BF16Vec8& v) {
+    float tmp_small[8];
+    RVVI(__riscv_vse32_v_f32, LMUL_256)(tmp_small, v.reg_fp32, 8);
+    float tmp_large[32];
+    for (int i = 0; i < 4; ++i) {
+      std::memcpy(tmp_large + (i * 8), tmp_small, 8 * sizeof(float));
+    }
+    reg_fp32 = RVVI(__riscv_vle32_v_f32, LMUL_1024)(tmp_large, 32);
+  }
+
+  void save(void* ptr) const {
+    float tmp[32];
+    RVVI(__riscv_vse32_v_f32, LMUL_1024)(tmp, reg_fp32, 32);
+    uint16_t* u16 = static_cast<uint16_t*>(ptr);
+    for (int i = 0; i < 32; ++i) {
+      uint32_t v;
+      std::memcpy(&v, &tmp[i], 4);
+      u16[i] = static_cast<uint16_t>(v >> 16);
+    }
+  }
+
+  void save(void* ptr, int elem_num) const {
+    float tmp[32];
+    RVVI(__riscv_vse32_v_f32, LMUL_1024)(tmp, reg_fp32, 32);
+    uint16_t* u16 = static_cast<uint16_t*>(ptr);
+    for (int i = 0; i < elem_num; ++i) {
+      uint32_t v;
+      std::memcpy(&v, &tmp[i], 4);
+      u16[i] = static_cast<uint16_t>(v >> 16);
+    }
+  }
+
+  void save_strided(void* ptr, ptrdiff_t stride) const {
+    float tmp[32];
+    RVVI(__riscv_vse32_v_f32, LMUL_1024)(tmp, reg_fp32, 32);
+    uint8_t* u8 = static_cast<uint8_t*>(ptr);
+    ptrdiff_t byte_stride = stride * sizeof(uint16_t);
+    for (int i = 0; i < 32; ++i) {
+      uint32_t v;
+      std::memcpy(&v, &tmp[i], 4);
+      uint16_t val = static_cast<uint16_t>(v >> 16);
+      *reinterpret_cast<uint16_t*>(u8 + i * byte_stride) = val;
+    }
+  }
+};
+#endif
+
+// ============================================================================
+// FP32 Implementation
+// ============================================================================
+
+struct FP32Vec4 : public Vec<FP32Vec4> {
+  constexpr static int VEC_ELEM_NUM = 4;
+  fixed_fp32x4_t reg;
+  explicit FP32Vec4(float v)
+      : reg(RVVI(__riscv_vfmv_v_f_f32, LMUL_128)(v, VEC_ELEM_NUM)) {};
+  explicit FP32Vec4()
+      : reg(RVVI(__riscv_vfmv_v_f_f32, LMUL_128)(0.0f, VEC_ELEM_NUM)) {};
+  explicit FP32Vec4(const float* ptr)
+      : reg(RVVI(__riscv_vle32_v_f32, LMUL_128)(ptr, VEC_ELEM_NUM)) {};
+  explicit FP32Vec4(fixed_fp32x4_t data) : reg(data) {};
+  explicit FP32Vec4(const FP32Vec4& data) : reg(data.reg) {};
+  void save(float* ptr) const {
+    RVVI(__riscv_vse32_v_f32, LMUL_128)(ptr, reg, VEC_ELEM_NUM);
+  }
+  void save(float* ptr, int elem_num) const {
+    RVVI(__riscv_vse32_v_f32, LMUL_128)(ptr, reg, elem_num);
+  }
+};
+
+struct FP32Vec8 : public Vec<FP32Vec8> {
+  constexpr static int VEC_ELEM_NUM = 8;
+  fixed_fp32x8_t reg;
+
+  explicit FP32Vec8(float v)
+      : reg(RVVI(__riscv_vfmv_v_f_f32, LMUL_256)(v, VEC_ELEM_NUM)) {};
+  explicit FP32Vec8()
+      : reg(RVVI(__riscv_vfmv_v_f_f32, LMUL_256)(0.0f, VEC_ELEM_NUM)) {};
+  explicit FP32Vec8(const float* ptr)
+      : reg(RVVI(__riscv_vle32_v_f32, LMUL_256)(ptr, VEC_ELEM_NUM)) {};
+  explicit FP32Vec8(fixed_fp32x8_t data) : reg(data) {};
+  explicit FP32Vec8(const FP32Vec8& data) : reg(data.reg) {};
+  explicit FP32Vec8(const FP16Vec8& v)
+      : reg(RVVI(__riscv_vfwcvt_f_f_v_f32, LMUL_256)(v.reg, VEC_ELEM_NUM)) {};
+  explicit FP32Vec8(fixed_fp16x8_t v)
+      : reg(RVVI(__riscv_vfwcvt_f_f_v_f32, LMUL_256)(v, VEC_ELEM_NUM)) {};
+
+#ifdef RISCV_BF16_SUPPORT
+  explicit FP32Vec8(fixed_bf16x8_t v)
+      : reg(RVVI(__riscv_vfwcvtbf16_f_f_v_f32, LMUL_256)(v, VEC_ELEM_NUM)) {};
+  explicit FP32Vec8(const BF16Vec8& v)
+      : reg(RVVI(__riscv_vfwcvtbf16_f_f_v_f32, LMUL_256)(v.reg, VEC_ELEM_NUM)) {
+        };
+#else
+  explicit FP32Vec8(const BF16Vec8& v) : reg(v.reg_fp32) {};
+#endif
+
+  float reduce_sum() const {
+    rvv_f32_accum_t scalar = __riscv_vfmv_s_f_f32m1(0.0f, 1);
+    scalar = RVVI3(__riscv_vfredusum_vs_f32, LMUL_256, _f32m1)(reg, scalar,
+                                                               VEC_ELEM_NUM);
+    return __riscv_vfmv_f_s_f32m1_f32(scalar);
+  }
+
+  FP32Vec8 operator*(const FP32Vec8& b) const {
+    return FP32Vec8(
+        RVVI(__riscv_vfmul_vv_f32, LMUL_256)(reg, b.reg, VEC_ELEM_NUM));
+  }
+  FP32Vec8 operator+(const FP32Vec8& b) const {
+    return FP32Vec8(
+        RVVI(__riscv_vfadd_vv_f32, LMUL_256)(reg, b.reg, VEC_ELEM_NUM));
+  }
+  FP32Vec8 operator-(const FP32Vec8& b) const {
+    return FP32Vec8(
+        RVVI(__riscv_vfsub_vv_f32, LMUL_256)(reg, b.reg, VEC_ELEM_NUM));
+  }
+  FP32Vec8 operator/(const FP32Vec8& b) const {
+    return FP32Vec8(
+        RVVI(__riscv_vfdiv_vv_f32, LMUL_256)(reg, b.reg, VEC_ELEM_NUM));
+  }
+
+  FP32Vec8 min(const FP32Vec8& b) const {
+    return FP32Vec8(
+        RVVI(__riscv_vfmin_vv_f32, LMUL_256)(reg, b.reg, VEC_ELEM_NUM));
+  }
+  FP32Vec8 max(const FP32Vec8& b) const {
+    return FP32Vec8(
+        RVVI(__riscv_vfmax_vv_f32, LMUL_256)(reg, b.reg, VEC_ELEM_NUM));
+  }
+  FP32Vec8 abs() const {
+    return FP32Vec8(RVVI(__riscv_vfabs_v_f32, LMUL_256)(reg, VEC_ELEM_NUM));
+  }
+
+  FP32Vec8 min(const FP32Vec8& b, int elem_num) const {
+    return FP32Vec8(RVVI(__riscv_vfmin_vv_f32, LMUL_256)(reg, b.reg, elem_num));
+  }
+  FP32Vec8 max(const FP32Vec8& b, int elem_num) const {
+    return FP32Vec8(RVVI(__riscv_vfmax_vv_f32, LMUL_256)(reg, b.reg, elem_num));
+  }
+
+  FP32Vec8 clamp(const FP32Vec8& min_v, const FP32Vec8& max_v) const {
+    fixed_fp32x8_t temp =
+        RVVI(__riscv_vfmax_vv_f32, LMUL_256)(min_v.reg, reg, VEC_ELEM_NUM);
+    return FP32Vec8(
+        RVVI(__riscv_vfmin_vv_f32, LMUL_256)(max_v.reg, temp, VEC_ELEM_NUM));
+  }
+
+  void save(float* ptr) const {
+    RVVI(__riscv_vse32_v_f32, LMUL_256)(ptr, reg, VEC_ELEM_NUM);
+  }
+  void save(float* ptr, int elem_num) const {
+    RVVI(__riscv_vse32_v_f32, LMUL_256)(ptr, reg, elem_num);
+  }
+  void save_strided(float* ptr, ptrdiff_t stride) const {
+    ptrdiff_t byte_stride = stride * sizeof(float);
+    RVVI(__riscv_vsse32_v_f32, LMUL_256)(ptr, byte_stride, reg, VEC_ELEM_NUM);
+  }
+
+  FP32Vec8 exp() const {
+    // Clamp input to prevent NaN: exp(-inf) must return 0, not NaN.
+    // Without clamping, -inf * 0.0 = NaN in the final poly * scale step.
+    // Matches the clamping strategy used by x86 AVX-512 and ARM NEON.
+    constexpr float exp_lo = -87.3365447505f;  // ln(FLT_MIN)
+    constexpr float exp_hi = 88.7228391117f;   // ln(FLT_MAX)
+    fixed_fp32x8_t x = RVVI(__riscv_vfmin_vf_f32, LMUL_256)(
+        RVVI(__riscv_vfmax_vf_f32, LMUL_256)(reg, exp_lo, VEC_ELEM_NUM), exp_hi,
+        VEC_ELEM_NUM);
+
+    const float inv_ln2 = 1.44269504088896341f;
+    fixed_fp32x8_t x_scaled =
+        RVVI(__riscv_vfmul_vf_f32, LMUL_256)(x, inv_ln2, VEC_ELEM_NUM);
+    fixed_i32x8_t n_int =
+        RVVI(__riscv_vfcvt_x_f_v_i32, LMUL_256)(x_scaled, VEC_ELEM_NUM);
+    fixed_fp32x8_t n_float =
+        RVVI(__riscv_vfcvt_f_x_v_f32, LMUL_256)(n_int, VEC_ELEM_NUM);
+
+    fixed_fp32x8_t r =
+        RVVI(__riscv_vfsub_vv_f32, LMUL_256)(x_scaled, n_float, VEC_ELEM_NUM);
+
+    fixed_fp32x8_t poly =
+        RVVI(__riscv_vfmv_v_f_f32, LMUL_256)(0.001333355810164f, VEC_ELEM_NUM);
+    poly = RVVI(__riscv_vfmul_vv_f32, LMUL_256)(poly, r, VEC_ELEM_NUM);
+    poly = RVVI(__riscv_vfadd_vf_f32, LMUL_256)(poly, 0.009618129107628f,
+                                                VEC_ELEM_NUM);
+    poly = RVVI(__riscv_vfmul_vv_f32, LMUL_256)(poly, r, VEC_ELEM_NUM);
+    poly = RVVI(__riscv_vfadd_vf_f32, LMUL_256)(poly, 0.055504108664821f,
+                                                VEC_ELEM_NUM);
+    poly = RVVI(__riscv_vfmul_vv_f32, LMUL_256)(poly, r, VEC_ELEM_NUM);
+    poly = RVVI(__riscv_vfadd_vf_f32, LMUL_256)(poly, 0.240226506959101f,
+                                                VEC_ELEM_NUM);
+    poly = RVVI(__riscv_vfmul_vv_f32, LMUL_256)(poly, r, VEC_ELEM_NUM);
+    poly = RVVI(__riscv_vfadd_vf_f32, LMUL_256)(poly, 0.693147180559945f,
+                                                VEC_ELEM_NUM);
+    poly = RVVI(__riscv_vfmul_vv_f32, LMUL_256)(poly, r, VEC_ELEM_NUM);
+    poly = RVVI(__riscv_vfadd_vf_f32, LMUL_256)(poly, 1.0f, VEC_ELEM_NUM);
+
+    fixed_i32x8_t biased_exp =
+        RVVI(__riscv_vadd_vx_i32, LMUL_256)(n_int, 127, VEC_ELEM_NUM);
+    biased_exp =
+        RVVI(__riscv_vmax_vx_i32, LMUL_256)(biased_exp, 0, VEC_ELEM_NUM);
+    fixed_i32x8_t exponent_bits =
+        RVVI(__riscv_vsll_vx_i32, LMUL_256)(biased_exp, 23, VEC_ELEM_NUM);
+    fixed_fp32x8_t scale = RVVI4(__riscv_vreinterpret_v_i32, LMUL_256, _f32,
+                                 LMUL_256)(exponent_bits);
+
+    return FP32Vec8(
+        RVVI(__riscv_vfmul_vv_f32, LMUL_256)(poly, scale, VEC_ELEM_NUM));
+  }
+
+  FP32Vec8 tanh() const {
+    fixed_fp32x8_t x_clamped = RVVI(__riscv_vfmin_vf_f32, LMUL_256)(
+        RVVI(__riscv_vfmax_vf_f32, LMUL_256)(reg, -9.0f, VEC_ELEM_NUM), 9.0f,
+        VEC_ELEM_NUM);
+    fixed_fp32x8_t x2 =
+        RVVI(__riscv_vfmul_vf_f32, LMUL_256)(x_clamped, 2.0f, VEC_ELEM_NUM);
+    FP32Vec8 exp_val = FP32Vec8(x2).exp();
+    fixed_fp32x8_t num =
+        RVVI(__riscv_vfsub_vf_f32, LMUL_256)(exp_val.reg, 1.0f, VEC_ELEM_NUM);
+    fixed_fp32x8_t den =
+        RVVI(__riscv_vfadd_vf_f32, LMUL_256)(exp_val.reg, 1.0f, VEC_ELEM_NUM);
+    return FP32Vec8(
+        RVVI(__riscv_vfdiv_vv_f32, LMUL_256)(num, den, VEC_ELEM_NUM));
+  }
+
+  FP32Vec8 er() const {
+    const float p = 0.3275911f, a1 = 0.254829592f, a2 = -0.284496736f,
+                a3 = 1.421413741f, a4 = -1.453152027f, a5 = 1.061405429f;
+    fixed_fp32x8_t abs_x =
+        RVVI(__riscv_vfabs_v_f32, LMUL_256)(reg, VEC_ELEM_NUM);
+
+    fixed_fp32x8_t t = RVVI(__riscv_vfadd_vf_f32, LMUL_256)(
+        RVVI(__riscv_vfmul_vf_f32, LMUL_256)(abs_x, p, VEC_ELEM_NUM), 1.0f,
+        VEC_ELEM_NUM);
+    t = RVVI(__riscv_vfrdiv_vf_f32, LMUL_256)(t, 1.0f, VEC_ELEM_NUM);
+
+    fixed_fp32x8_t poly =
+        RVVI(__riscv_vfmv_v_f_f32, LMUL_256)(a5, VEC_ELEM_NUM);
+    poly = RVVI(__riscv_vfadd_vf_f32, LMUL_256)(
+        RVVI(__riscv_vfmul_vv_f32, LMUL_256)(poly, t, VEC_ELEM_NUM), a4,
+        VEC_ELEM_NUM);
+    poly = RVVI(__riscv_vfadd_vf_f32, LMUL_256)(
+        RVVI(__riscv_vfmul_vv_f32, LMUL_256)(poly, t, VEC_ELEM_NUM), a3,
+        VEC_ELEM_NUM);
+    poly = RVVI(__riscv_vfadd_vf_f32, LMUL_256)(
+        RVVI(__riscv_vfmul_vv_f32, LMUL_256)(poly, t, VEC_ELEM_NUM), a2,
+        VEC_ELEM_NUM);
+    poly = RVVI(__riscv_vfadd_vf_f32, LMUL_256)(
+        RVVI(__riscv_vfmul_vv_f32, LMUL_256)(poly, t, VEC_ELEM_NUM), a1,
+        VEC_ELEM_NUM);
+    poly = RVVI(__riscv_vfmul_vv_f32, LMUL_256)(poly, t, VEC_ELEM_NUM);
+
+    fixed_fp32x8_t exp_val = FP32Vec8(RVVI(__riscv_vfneg_v_f32, LMUL_256)(
+                                          RVVI(__riscv_vfmul_vv_f32, LMUL_256)(
+                                              abs_x, abs_x, VEC_ELEM_NUM),
+                                          VEC_ELEM_NUM))
+                                 .exp()
+                                 .reg;
+    fixed_fp32x8_t res = RVVI(__riscv_vfrsub_vf_f32, LMUL_256)(
+        RVVI(__riscv_vfmul_vv_f32, LMUL_256)(poly, exp_val, VEC_ELEM_NUM), 1.0f,
+        VEC_ELEM_NUM);
+
+    rvv_mask_f32x8_t mask = RVVIB(__riscv_vmflt_vf_f32, LMUL_256, BOOL_256)(
+        reg, 0.0f, VEC_ELEM_NUM);
+    return FP32Vec8(
+        RVVI3(__riscv_vfneg_v_f32, LMUL_256, _m)(mask, res, VEC_ELEM_NUM));
+  }
+};
+
+struct FP32Vec16 : public Vec<FP32Vec16> {
+  constexpr static int VEC_ELEM_NUM = 16;
+  fixed_fp32x16_t reg;
+
+  explicit FP32Vec16(float v)
+      : reg(RVVI(__riscv_vfmv_v_f_f32, LMUL_512)(v, VEC_ELEM_NUM)) {};
+  explicit FP32Vec16()
+      : reg(RVVI(__riscv_vfmv_v_f_f32, LMUL_512)(0.0f, VEC_ELEM_NUM)) {};
+  explicit FP32Vec16(const float* ptr)
+      : reg(RVVI(__riscv_vle32_v_f32, LMUL_512)(ptr, VEC_ELEM_NUM)) {};
+  explicit FP32Vec16(fixed_fp32x16_t data) : reg(data) {};
+  explicit FP32Vec16(const FP32Vec8& data)
+      : reg(RVVI4(__riscv_vcreate_v_f32, LMUL_256, _f32, LMUL_512)(
+            data.reg, data.reg)) {};
+  explicit FP32Vec16(const FP32Vec16& data) : reg(data.reg) {};
+  explicit FP32Vec16(const FP16Vec16& v);
+
+#ifdef RISCV_BF16_SUPPORT
+  explicit FP32Vec16(fixed_bf16x16_t v)
+      : reg(RVVI(__riscv_vfwcvtbf16_f_f_v_f32, LMUL_512)(v, VEC_ELEM_NUM)) {};
+  explicit FP32Vec16(const BF16Vec16& v)
+      : reg(RVVI(__riscv_vfwcvtbf16_f_f_v_f32, LMUL_512)(v.reg, VEC_ELEM_NUM)) {
+        };
+#else
+  explicit FP32Vec16(const BF16Vec16& v) : reg(v.reg_fp32) {};
+#endif
+
+  FP32Vec16 operator+(const FP32Vec16& b) const {
+    return FP32Vec16(
+        RVVI(__riscv_vfadd_vv_f32, LMUL_512)(reg, b.reg, VEC_ELEM_NUM));
+  }
+  FP32Vec16 operator-(const FP32Vec16& b) const {
+    return FP32Vec16(
+        RVVI(__riscv_vfsub_vv_f32, LMUL_512)(reg, b.reg, VEC_ELEM_NUM));
+  }
+  FP32Vec16 operator*(const FP32Vec16& b) const {
+    return FP32Vec16(
+        RVVI(__riscv_vfmul_vv_f32, LMUL_512)(reg, b.reg, VEC_ELEM_NUM));
+  }
+  FP32Vec16 operator/(const FP32Vec16& b) const {
+    return FP32Vec16(
+        RVVI(__riscv_vfdiv_vv_f32, LMUL_512)(reg, b.reg, VEC_ELEM_NUM));
+  }
+
+  FP32Vec16 fma(const FP32Vec16& a, const FP32Vec16& b) const {
+    return FP32Vec16(
+        RVVI(__riscv_vfmacc_vv_f32, LMUL_512)(reg, a.reg, b.reg, VEC_ELEM_NUM));
+  }
+
+  float reduce_sum() const {
+    rvv_f32_accum_t scalar = __riscv_vfmv_s_f_f32m1(0.0f, 1);
+    scalar = RVVI3(__riscv_vfredusum_vs_f32, LMUL_512, _f32m1)(reg, scalar,
+                                                               VEC_ELEM_NUM);
+    return __riscv_vfmv_f_s_f32m1_f32(scalar);
+  }
+
+  float reduce_max() const {
+    rvv_f32_accum_t scalar =
+        __riscv_vfmv_s_f_f32m1(std::numeric_limits<float>::lowest(), 1);
+    scalar = RVVI3(__riscv_vfredmax_vs_f32, LMUL_512, _f32m1)(reg, scalar,
+                                                              VEC_ELEM_NUM);
+    return __riscv_vfmv_f_s_f32m1_f32(scalar);
+  }
+
+  float reduce_min() const {
+    rvv_f32_accum_t scalar =
+        __riscv_vfmv_s_f_f32m1(std::numeric_limits<float>::max(), 1);
+    scalar = RVVI3(__riscv_vfredmin_vs_f32, LMUL_512, _f32m1)(reg, scalar,
+                                                              VEC_ELEM_NUM);
+    return __riscv_vfmv_f_s_f32m1_f32(scalar);
+  }
+
+  template <int group_size>
+  float reduce_sub_sum(int idx) {
+    static_assert(VEC_ELEM_NUM % group_size == 0);
+    const int start = idx * group_size;
+    auto indices = RVVI(__riscv_vid_v_u32, LMUL_512)(VEC_ELEM_NUM);
+    rvv_mask_f32x16_t mask = RVVI(__riscv_vmand_mm_, BOOL_512)(
+        RVVIB(__riscv_vmsgeu_vx_u32, LMUL_512, BOOL_512)(indices, start,
+                                                         VEC_ELEM_NUM),
+        RVVIB(__riscv_vmsltu_vx_u32, LMUL_512, BOOL_512)(
+            indices, start + group_size, VEC_ELEM_NUM),
+        VEC_ELEM_NUM);
+    rvv_f32_accum_t scalar = __riscv_vfmv_s_f_f32m1(0.0f, 1);
+    scalar = RVVI3(__riscv_vfredusum_vs_f32, LMUL_512, _f32m1_m)(
+        mask, reg, scalar, VEC_ELEM_NUM);
+    return __riscv_vfmv_f_s_f32m1_f32(scalar);
+  };
+
+  FP32Vec16 max(const FP32Vec16& b) const {
+    return FP32Vec16(
+        RVVI(__riscv_vfmax_vv_f32, LMUL_512)(reg, b.reg, VEC_ELEM_NUM));
+  }
+  FP32Vec16 min(const FP32Vec16& b) const {
+    return FP32Vec16(
+        RVVI(__riscv_vfmin_vv_f32, LMUL_512)(reg, b.reg, VEC_ELEM_NUM));
+  }
+  FP32Vec16 abs() const {
+    return FP32Vec16(RVVI(__riscv_vfabs_v_f32, LMUL_512)(reg, VEC_ELEM_NUM));
+  }
+
+  FP32Vec16 clamp(const FP32Vec16& min_v, const FP32Vec16& max_v) const {
+    return FP32Vec16(RVVI(__riscv_vfmin_vv_f32, LMUL_512)(
+        max_v.reg,
+        RVVI(__riscv_vfmax_vv_f32, LMUL_512)(min_v.reg, reg, VEC_ELEM_NUM),
+        VEC_ELEM_NUM));
+  }
+
+  void save(float* ptr) const {
+    RVVI(__riscv_vse32_v_f32, LMUL_512)(ptr, reg, VEC_ELEM_NUM);
+  }
+  void save(float* ptr, int elem_num) const {
+    RVVI(__riscv_vse32_v_f32, LMUL_512)(ptr, reg, elem_num);
+  }
+  void save_strided(float* ptr, ptrdiff_t stride) const {
+    ptrdiff_t byte_stride = stride * sizeof(float);
+    RVVI(__riscv_vsse32_v_f32, LMUL_512)(ptr, byte_stride, reg, VEC_ELEM_NUM);
+  }
+
+  FP32Vec16 exp() const {
+    // Clamp input to prevent NaN: exp(-inf) must return 0, not NaN.
+    // Without clamping, -inf * 0.0 = NaN in the final poly * scale step.
+    // Matches the clamping strategy used by x86 AVX-512 and ARM NEON.
+    constexpr float exp_lo = -87.3365447505f;  // ln(FLT_MIN)
+    constexpr float exp_hi = 88.7228391117f;   // ln(FLT_MAX)
+    fixed_fp32x16_t x = RVVI(__riscv_vfmin_vf_f32, LMUL_512)(
+        RVVI(__riscv_vfmax_vf_f32, LMUL_512)(reg, exp_lo, VEC_ELEM_NUM), exp_hi,
+        VEC_ELEM_NUM);
+
+    const float inv_ln2 = 1.44269504088896341f;
+    fixed_fp32x16_t x_scaled =
+        RVVI(__riscv_vfmul_vf_f32, LMUL_512)(x, inv_ln2, VEC_ELEM_NUM);
+    fixed_i32x16_t n_int =
+        RVVI(__riscv_vfcvt_x_f_v_i32, LMUL_512)(x_scaled, VEC_ELEM_NUM);
+    fixed_fp32x16_t n_float =
+        RVVI(__riscv_vfcvt_f_x_v_f32, LMUL_512)(n_int, VEC_ELEM_NUM);
+    fixed_fp32x16_t r =
+        RVVI(__riscv_vfsub_vv_f32, LMUL_512)(x_scaled, n_float, VEC_ELEM_NUM);
+
+    fixed_fp32x16_t poly =
+        RVVI(__riscv_vfmv_v_f_f32, LMUL_512)(0.001333355810164f, VEC_ELEM_NUM);
+    poly = RVVI(__riscv_vfadd_vf_f32, LMUL_512)(
+        RVVI(__riscv_vfmul_vv_f32, LMUL_512)(poly, r, VEC_ELEM_NUM),
+        0.009618129107628f, VEC_ELEM_NUM);
+    poly = RVVI(__riscv_vfadd_vf_f32, LMUL_512)(
+        RVVI(__riscv_vfmul_vv_f32, LMUL_512)(poly, r, VEC_ELEM_NUM),
+        0.055504108664821f, VEC_ELEM_NUM);
+    poly = RVVI(__riscv_vfadd_vf_f32, LMUL_512)(
+        RVVI(__riscv_vfmul_vv_f32, LMUL_512)(poly, r, VEC_ELEM_NUM),
+        0.240226506959101f, VEC_ELEM_NUM);
+    poly = RVVI(__riscv_vfadd_vf_f32, LMUL_512)(
+        RVVI(__riscv_vfmul_vv_f32, LMUL_512)(poly, r, VEC_ELEM_NUM),
+        0.693147180559945f, VEC_ELEM_NUM);
+    poly = RVVI(__riscv_vfadd_vf_f32, LMUL_512)(
+        RVVI(__riscv_vfmul_vv_f32, LMUL_512)(poly, r, VEC_ELEM_NUM), 1.0f,
+        VEC_ELEM_NUM);
+
+    fixed_i32x16_t biased_exp = RVVI(__riscv_vmax_vx_i32, LMUL_512)(
+        RVVI(__riscv_vadd_vx_i32, LMUL_512)(n_int, 127, VEC_ELEM_NUM), 0,
+        VEC_ELEM_NUM);
+    fixed_fp32x16_t scale =
+        RVVI4(__riscv_vreinterpret_v_i32, LMUL_512, _f32, LMUL_512)(
+            RVVI(__riscv_vsll_vx_i32, LMUL_512)(biased_exp, 23, VEC_ELEM_NUM));
+
+    return FP32Vec16(
+        RVVI(__riscv_vfmul_vv_f32, LMUL_512)(poly, scale, VEC_ELEM_NUM));
+  }
+
+  FP32Vec16 tanh() const {
+    fixed_fp32x16_t x_clamped = RVVI(__riscv_vfmin_vf_f32, LMUL_512)(
+        RVVI(__riscv_vfmax_vf_f32, LMUL_512)(reg, -9.0f, VEC_ELEM_NUM), 9.0f,
+        VEC_ELEM_NUM);
+    FP32Vec16 exp_val = FP32Vec16(RVVI(__riscv_vfmul_vf_f32, LMUL_512)(
+                                      x_clamped, 2.0f, VEC_ELEM_NUM))
+                            .exp();
+    return FP32Vec16(RVVI(__riscv_vfdiv_vv_f32, LMUL_512)(
+        RVVI(__riscv_vfsub_vf_f32, LMUL_512)(exp_val.reg, 1.0f, VEC_ELEM_NUM),
+        RVVI(__riscv_vfadd_vf_f32, LMUL_512)(exp_val.reg, 1.0f, VEC_ELEM_NUM),
+        VEC_ELEM_NUM));
+  }
+
+  FP32Vec16 er() const {
+    const float p = 0.3275911f, a1 = 0.254829592f, a2 = -0.284496736f,
+                a3 = 1.421413741f, a4 = -1.453152027f, a5 = 1.061405429f;
+    fixed_fp32x16_t abs_x =
+        RVVI(__riscv_vfabs_v_f32, LMUL_512)(reg, VEC_ELEM_NUM);
+    fixed_fp32x16_t t = RVVI(__riscv_vfrdiv_vf_f32, LMUL_512)(
+        RVVI(__riscv_vfadd_vf_f32, LMUL_512)(
+            RVVI(__riscv_vfmul_vf_f32, LMUL_512)(abs_x, p, VEC_ELEM_NUM), 1.0f,
+            VEC_ELEM_NUM),
+        1.0f, VEC_ELEM_NUM);
+
+    fixed_fp32x16_t poly =
+        RVVI(__riscv_vfmv_v_f_f32, LMUL_512)(a5, VEC_ELEM_NUM);
+    poly = RVVI(__riscv_vfadd_vf_f32, LMUL_512)(
+        RVVI(__riscv_vfmul_vv_f32, LMUL_512)(poly, t, VEC_ELEM_NUM), a4,
+        VEC_ELEM_NUM);
+    poly = RVVI(__riscv_vfadd_vf_f32, LMUL_512)(
+        RVVI(__riscv_vfmul_vv_f32, LMUL_512)(poly, t, VEC_ELEM_NUM), a3,
+        VEC_ELEM_NUM);
+    poly = RVVI(__riscv_vfadd_vf_f32, LMUL_512)(
+        RVVI(__riscv_vfmul_vv_f32, LMUL_512)(poly, t, VEC_ELEM_NUM), a2,
+        VEC_ELEM_NUM);
+    poly = RVVI(__riscv_vfadd_vf_f32, LMUL_512)(
+        RVVI(__riscv_vfmul_vv_f32, LMUL_512)(poly, t, VEC_ELEM_NUM), a1,
+        VEC_ELEM_NUM);
+    poly = RVVI(__riscv_vfmul_vv_f32, LMUL_512)(poly, t, VEC_ELEM_NUM);
+
+    fixed_fp32x16_t exp_val =
+        FP32Vec16(RVVI(__riscv_vfneg_v_f32, LMUL_512)(
+                      RVVI(__riscv_vfmul_vv_f32, LMUL_512)(abs_x, abs_x,
+                                                           VEC_ELEM_NUM),
+                      VEC_ELEM_NUM))
+            .exp()
+            .reg;
+    fixed_fp32x16_t res = RVVI(__riscv_vfrsub_vf_f32, LMUL_512)(
+        RVVI(__riscv_vfmul_vv_f32, LMUL_512)(poly, exp_val, VEC_ELEM_NUM), 1.0f,
+        VEC_ELEM_NUM);
+
+    rvv_mask_f32x16_t mask = RVVIB(__riscv_vmflt_vf_f32, LMUL_512, BOOL_512)(
+        reg, 0.0f, VEC_ELEM_NUM);
+    return FP32Vec16(
+        RVVI3(__riscv_vfneg_v_f32, LMUL_512, _m)(mask, res, VEC_ELEM_NUM));
+  }
+};
+
+// ============================================================================
+// Type Traits & Global Helpers
+// ============================================================================
+
+template <typename T>
+struct VecType {
+  using vec_type = void;
+  using vec_t = void;
+};
+
+template <typename T>
+using vec_t = typename VecType<T>::vec_type;
+
+template <>
+struct VecType<float> {
+  using vec_type = FP32Vec8;
+  using vec_t = FP32Vec8;
+};
+template <>
+struct VecType<c10::Half> {
+  using vec_type = FP16Vec8;
+  using vec_t = FP16Vec8;
+};
+template <>
+struct VecType<c10::BFloat16> {
+  using vec_type = BF16Vec8;
+  using vec_t = BF16Vec8;
+};
+
+template <typename T>
+void storeFP32(float v, T* ptr) {
+  *ptr = v;
+}
+template <>
+inline void storeFP32<c10::Half>(float v, c10::Half* ptr) {
+  *reinterpret_cast<_Float16*>(ptr) = static_cast<_Float16>(v);
+}
+
+inline FP16Vec16::FP16Vec16(const FP32Vec16& v) {
+  reg = RVVI(__riscv_vfncvt_f_f_w_f16, LMUL_256)(v.reg, VEC_ELEM_NUM);
+}
+inline FP16Vec8::FP16Vec8(const FP32Vec8& v) {
+  reg = RVVI(__riscv_vfncvt_f_f_w_f16, LMUL_128)(v.reg, VEC_ELEM_NUM);
+}
+inline FP32Vec16::FP32Vec16(const FP16Vec16& v) {
+  reg = RVVI(__riscv_vfwcvt_f_f_v_f32, LMUL_512)(v.reg, VEC_ELEM_NUM);
+}
+inline void fma(FP32Vec16& acc, const FP32Vec16& a, const FP32Vec16& b) {
+  acc = acc.fma(a, b);
+}
+
+#ifdef RISCV_BF16_SUPPORT
+template <>
+inline void storeFP32<c10::BFloat16>(float v, c10::BFloat16* ptr) {
+  *ptr = static_cast<__bf16>(v);
+};
+inline BF16Vec8::BF16Vec8(const FP32Vec8& v)
+    : reg(RVVI(__riscv_vfncvtbf16_f_f_w_bf16, LMUL_128)(v.reg, VEC_ELEM_NUM)) {
+      };
+inline BF16Vec16::BF16Vec16(const FP32Vec16& v)
+    : reg(RVVI(__riscv_vfncvtbf16_f_f_w_bf16, LMUL_256)(v.reg, VEC_ELEM_NUM)) {
+      };
+#else
+template <>
+inline void storeFP32<c10::BFloat16>(float v, c10::BFloat16* ptr) {
+  uint32_t val;
+  std::memcpy(&val, &v, 4);
+  *reinterpret_cast<uint16_t*>(ptr) = static_cast<uint16_t>(val >> 16);
+}
+inline BF16Vec8::BF16Vec8(const FP32Vec8& v) : reg_fp32(v.reg) {}
+inline BF16Vec16::BF16Vec16(const FP32Vec16& v) : reg_fp32(v.reg) {}
+#endif
+
+inline void prefetch(const void* addr) { __builtin_prefetch(addr, 0, 1); }
+
+}  // namespace vec_op
+
+#ifndef CPU_KERNEL_GUARD_IN
+  #define CPU_KERNEL_GUARD_IN(NAME)
+#endif
+
+#ifndef CPU_KERNEL_GUARD_OUT
+  #define CPU_KERNEL_GUARD_OUT(NAME)
+#endif
+
+#endif  // CPU_TYPES_RISCV_IMPL_HPP

csrc/cpu/pos_encoding.cpp

@@ -178,7 +178,12 @@ void rotary_embedding_gptj_impl(
 
 void rotary_embedding(torch::Tensor& positions, torch::Tensor& query,
                       std::optional<torch::Tensor> key, int64_t head_size,
-                      torch::Tensor& cos_sin_cache, bool is_neox) {
+                      torch::Tensor& cos_sin_cache, bool is_neox,
+                      int64_t rope_dim_offset, bool inverse) {
+  TORCH_CHECK(rope_dim_offset == 0,
+              "rope_dim_offset != 0 is not supported on CPU");
+  TORCH_CHECK(!inverse, "inverse rotary embedding is not supported on CPU");
+
   int num_tokens = positions.numel();
   int rot_dim = cos_sin_cache.size(1);
   int num_heads = query.size(-1) / head_size;

csrc/cpu/torch_bindings.cpp

@@ -263,7 +263,8 @@ TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, ops) {
   ops.def(
       "rotary_embedding(Tensor positions, Tensor! query,"
       "                 Tensor!? key, int head_size,"
-      "                 Tensor cos_sin_cache, bool is_neox) -> ()");
+      "                 Tensor cos_sin_cache, bool is_neox, int "
+      "rope_dim_offset=0, bool inverse=False) -> ()");
   ops.impl("rotary_embedding", torch::kCPU, &rotary_embedding);
 
   // Quantization

csrc/fused_deepseek_v4_qnorm_rope_kv_insert_kernel.cu

@@ -0,0 +1,477 @@
+/*
+ * SPDX-License-Identifier: Apache-2.0
+ * SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+ *
+ * Horizontally-fused DeepseekV4-MLA kernel:
+ *   - Q side:  per-head RMSNorm (no weight) + GPT-J RoPE on last ROPE_DIM
+ *   - KV side: GPT-J RoPE on last ROPE_DIM + UE8M0 FP8 quant on NoPE + paged
+ *              cache insert
+ *
+ * Structured after `applyMLARopeAndAssignQKVKernelGeneration` in
+ * TensorRT-LLM's mlaKernels.cu: one kernel, one grid, with head-slot
+ * dispatch choosing Q vs KV work per warp.  The per-warp RMSNorm/RoPE
+ * skeleton is adapted from vllm-deepseek_v4's existing
+ * `fusedQKNormRopeKernel` (csrc/fused_qknorm_rope_kernel.cu).
+ *
+ * Assumptions (hard-coded for DeepseekV4 attention):
+ *   HEAD_DIM  = 512
+ *   ROPE_DIM  = 64   (RoPE applied to dims [NOPE_DIM, HEAD_DIM))
+ *   NOPE_DIM  = 448
+ *   QUANT_BLOCK = 64 (UE8M0 FP8 quant block)
+ *   FP8_MAX   = 448.0f
+ *   is_neox=false (GPT-J interleaved pairs)
+ *   cos_sin_cache layout [max_pos, rope_dim] = cos || sin (cos first, sin
+ *     second along last dim; each half is rope_dim/2 = 32 values)
+ *
+ * Cache layout per paged-cache block (block_size tokens):
+ *   [0,            bs*576):          token data, 448 fp8 + 128 bf16 each
+ *   [bs*576,       bs*576 + bs*8):   UE8M0 scales, 7 real + 1 pad per token
+ */
+
+#include <cmath>
+#include <cuda_fp8.h>
+#include <cuda_runtime.h>
+#include <type_traits>
+
+#include <ATen/cuda/CUDAContext.h>
+#include <c10/cuda/CUDAGuard.h>
+#include <torch/cuda.h>
+
+#include "cuda_compat.h"
+#include "dispatch_utils.h"
+#include "type_convert.cuh"
+
+#ifndef FINAL_MASK
+  #define FINAL_MASK 0xffffffffu
+#endif
+
+namespace vllm {
+namespace deepseek_v4_fused_ops {
+
+namespace {
+inline int getSMVersion() {
+  auto* props = at::cuda::getCurrentDeviceProperties();
+  return props->major * 10 + props->minor;
+}
+}  // namespace
+
+// ────────────────────────────────────────────────────────────────────────────
+// Constants
+// ────────────────────────────────────────────────────────────────────────────
+constexpr int kHeadDim = 512;
+constexpr int kRopeDim = 64;
+constexpr int kNopeDim = kHeadDim - kRopeDim;  // 448
+constexpr int kQuantBlock = 64;
+constexpr int kNumQuantBlocks = kNopeDim / kQuantBlock;   // 7
+constexpr int kScaleBytesPerToken = kNumQuantBlocks + 1;  // 8 (7 real + 1 pad)
+constexpr int kTokenDataBytes = kNopeDim + kRopeDim * 2;  // 448 + 128 = 576
+constexpr float kFp8Max = 448.0f;
+
+// Per-warp layout:  32 lanes × 16 elems/lane = 512 elems = HEAD_DIM.
+constexpr int kNumLanes = 32;
+constexpr int kElemsPerLane = kHeadDim / kNumLanes;  // 16
+
+// ────────────────────────────────────────────────────────────────────────────
+// Small inline helpers
+// ────────────────────────────────────────────────────────────────────────────
+__device__ __forceinline__ float warp4MaxAbs(float val) {
+  // Reduce absolute max across 4 consecutive lanes (lane id & 3 group).
+  float peer = __shfl_xor_sync(FINAL_MASK, val, 1);
+  val = fmaxf(val, peer);
+  peer = __shfl_xor_sync(FINAL_MASK, val, 2);
+  val = fmaxf(val, peer);
+  return val;
+}
+
+template <typename T>
+__device__ __forceinline__ float warpSum(float val) {
+#pragma unroll
+  for (int mask = 16; mask > 0; mask >>= 1) {
+    val += __shfl_xor_sync(FINAL_MASK, val, mask, 32);
+  }
+  return val;
+}
+
+// ────────────────────────────────────────────────────────────────────────────
+// Kernel
+// ────────────────────────────────────────────────────────────────────────────
+//
+// Grid: 1D, gridDim.x = ceil(num_tokens_full * (num_heads_q + 1) /
+// warps_per_block) Block: blockDim.x = 256 threads (8 warps per block) Each
+// warp handles one (token, head_slot) pair. head_slot < num_heads_q          →
+// Q branch (RMSNorm + RoPE, in place) head_slot == num_heads_q         → KV
+// branch (RoPE + UE8M0 quant + insert)
+//
+// With DP padding, q/kv/position_ids can have more rows than slot_mapping.
+// The Q branch covers all `num_tokens_full` rows (downstream attention uses
+// them).  The KV branch only inserts the first `num_tokens_insert` tokens
+// (= slot_mapping length) into the paged cache.
+//
+template <typename scalar_t_in>
+__global__ void fusedDeepseekV4QNormRopeKVRopeQuantInsertKernel(
+    scalar_t_in* __restrict__ q_inout,         // [N, H, 512] bf16, in place
+    scalar_t_in const* __restrict__ kv_in,     // [N, 512] bf16
+    uint8_t* __restrict__ k_cache,             // [num_blocks, block_stride]
+    int64_t const* __restrict__ slot_mapping,  // [num_tokens_insert] i64
+    int64_t const* __restrict__ position_ids,  // [N] i64
+    float const* __restrict__ cos_sin_cache,   // [max_pos, 64] fp32
+    float const eps,
+    int const num_tokens_full,    // = q.size(0) = kv.size(0)
+    int const num_tokens_insert,  // = slot_mapping.size(0), ≤ num_tokens_full
+    int const num_heads_q,        // H
+    int const cache_block_size,   // tokens per paged-cache block
+    int const kv_block_stride) {  // bytes per paged-cache block
+#if (!defined(__CUDA_ARCH__) || __CUDA_ARCH__ < 800) && !defined(USE_ROCM)
+  // BF16 _typeConvert specialization is unavailable on pre-Ampere.  The
+  // DeepseekV4 kernel only runs with bf16 inputs in practice, so compile a
+  // no-op stub for sm_70/sm_75 to keep multi-arch builds happy.
+  if constexpr (std::is_same_v<scalar_t_in, c10::BFloat16>) {
+    return;
+  } else {
+#endif
+    using Converter = vllm::_typeConvert<scalar_t_in>;
+
+    int const warpsPerBlock = blockDim.x / 32;
+    int const warpId = threadIdx.x / 32;
+    int const laneId = threadIdx.x % 32;
+    int const globalWarpIdx = blockIdx.x * warpsPerBlock + warpId;
+
+    int const total_slots_per_token = num_heads_q + 1;
+    int const tokenIdx = globalWarpIdx / total_slots_per_token;
+    int const slotIdx = globalWarpIdx % total_slots_per_token;
+    if (tokenIdx >= num_tokens_full) return;
+
+    bool const isKV = (slotIdx == num_heads_q);
+    // KV branch: skip DP-padded tokens (no slot reserved for them).
+    if (isKV && tokenIdx >= num_tokens_insert) return;
+
+    // PDL: wait for predecessor kernel (upstream q/kv producer) to signal
+    // before touching any global memory.  No-op when PDL is not enabled on
+    // the launch.  The CUDA runtime wrapper emits the griddepcontrol.wait
+    // PTX with the required memory clobber internally.
+#if defined(__CUDA_ARCH__) && (__CUDA_ARCH__ >= 900)
+    cudaGridDependencySynchronize();
+#endif
+
+    // Dim range this lane owns within the 512-wide head.
+    int const dim_base = laneId * kElemsPerLane;  // in [0, 512) step 16
+
+    // ── Load 16 bf16 → 16 fp32 registers (one 16-byte + one 16-byte LDG) ────
+    float elements[kElemsPerLane];
+    float sumOfSquares = 0.0f;
+
+    scalar_t_in const* src_ptr;
+    if (isKV) {
+      src_ptr = kv_in + static_cast<int64_t>(tokenIdx) * kHeadDim + dim_base;
+    } else {
+      int64_t const q_row_offset =
+          (static_cast<int64_t>(tokenIdx) * num_heads_q + slotIdx) * kHeadDim +
+          dim_base;
+      src_ptr = q_inout + q_row_offset;
+    }
+
+    // Two 16-byte loads per thread (8 bf16 each).  Use uint4 as the vector
+    // type and bitcast to scalar_t_in packed pairs for conversion.
+    uint4 v0 = *reinterpret_cast<uint4 const*>(src_ptr);
+    uint4 v1 = *reinterpret_cast<uint4 const*>(src_ptr + 8);
+
+    {
+      typename Converter::packed_hip_type const* p0 =
+          reinterpret_cast<typename Converter::packed_hip_type const*>(&v0);
+      typename Converter::packed_hip_type const* p1 =
+          reinterpret_cast<typename Converter::packed_hip_type const*>(&v1);
+// Each packed_hip_type holds 2 bf16 → 4 packed = 8 elems per uint4.
+#pragma unroll
+      for (int i = 0; i < 4; i++) {
+        float2 f2 = Converter::convert(p0[i]);
+        elements[2 * i] = f2.x;
+        elements[2 * i + 1] = f2.y;
+      }
+#pragma unroll
+      for (int i = 0; i < 4; i++) {
+        float2 f2 = Converter::convert(p1[i]);
+        elements[8 + 2 * i] = f2.x;
+        elements[8 + 2 * i + 1] = f2.y;
+      }
+    }
+
+    // ── Q branch: RMSNorm with no weight (has_weight=False) ─────────────────
+    // Variance + rsqrt + multiply all in fp32, no intermediate bf16 round.
+    // The downstream bf16 round only happens at the final store.
+    if (!isKV) {
+#pragma unroll
+      for (int i = 0; i < kElemsPerLane; i++) {
+        sumOfSquares += elements[i] * elements[i];
+      }
+      sumOfSquares = warpSum<float>(sumOfSquares);
+      float const rms_rcp =
+          rsqrtf(sumOfSquares / static_cast<float>(kHeadDim) + eps);
+#pragma unroll
+      for (int i = 0; i < kElemsPerLane; i++) {
+        elements[i] = elements[i] * rms_rcp;
+      }
+    }
+
+    // ── GPT-J RoPE on dims [NOPE_DIM, HEAD_DIM) ─────────────────────────────
+    // All math in fp32.  cos_sin_cache is loaded as fp32 (its native storage).
+    bool const is_rope_lane = dim_base >= kNopeDim;
+    if (is_rope_lane) {
+      int64_t const pos = position_ids[tokenIdx];
+      constexpr int kHalfRope = kRopeDim / 2;  // 32
+      float const* cos_ptr = cos_sin_cache + pos * kRopeDim;
+      float const* sin_ptr = cos_ptr + kHalfRope;
+
+      int const rope_local_base = dim_base - kNopeDim;  // in [0, 64) step 16
+#pragma unroll
+      for (int p = 0; p < kElemsPerLane / 2; p++) {
+        int const pair_dim = rope_local_base + 2 * p;
+        int const half_idx = pair_dim / 2;
+        float const cos_v = VLLM_LDG(cos_ptr + half_idx);
+        float const sin_v = VLLM_LDG(sin_ptr + half_idx);
+        float const x_even = elements[2 * p];
+        float const x_odd = elements[2 * p + 1];
+        elements[2 * p] = x_even * cos_v - x_odd * sin_v;
+        elements[2 * p + 1] = x_even * sin_v + x_odd * cos_v;
+      }
+    }
+
+    // ═══════════════════════════════════════════════════════════════════════
+    // Q branch: cast to bf16 and store back in place.
+    // ═══════════════════════════════════════════════════════════════════════
+    if (!isKV) {
+      uint4 out0, out1;
+      typename Converter::packed_hip_type* po0 =
+          reinterpret_cast<typename Converter::packed_hip_type*>(&out0);
+      typename Converter::packed_hip_type* po1 =
+          reinterpret_cast<typename Converter::packed_hip_type*>(&out1);
+#pragma unroll
+      for (int i = 0; i < 4; i++) {
+        po0[i] = Converter::convert(
+            make_float2(elements[2 * i], elements[2 * i + 1]));
+      }
+#pragma unroll
+      for (int i = 0; i < 4; i++) {
+        po1[i] = Converter::convert(
+            make_float2(elements[8 + 2 * i], elements[8 + 2 * i + 1]));
+      }
+      scalar_t_in* dst =
+          q_inout +
+          (static_cast<int64_t>(tokenIdx) * num_heads_q + slotIdx) * kHeadDim +
+          dim_base;
+      *reinterpret_cast<uint4*>(dst) = out0;
+      *reinterpret_cast<uint4*>(dst + 8) = out1;
+#if defined(__CUDA_ARCH__) && (__CUDA_ARCH__ >= 900)
+      cudaTriggerProgrammaticLaunchCompletion();
+#endif
+      return;
+    }
+
+    // ═══════════════════════════════════════════════════════════════════════
+    // KV branch.
+    // ═══════════════════════════════════════════════════════════════════════
+    int64_t const slot_id = slot_mapping[tokenIdx];
+    if (slot_id < 0) {
+#if defined(__CUDA_ARCH__) && (__CUDA_ARCH__ >= 900)
+      cudaTriggerProgrammaticLaunchCompletion();
+#endif
+      return;
+    }
+
+    int64_t const block_idx = slot_id / cache_block_size;
+    int64_t const pos_in_block = slot_id % cache_block_size;
+    uint8_t* block_base =
+        k_cache + block_idx * static_cast<int64_t>(kv_block_stride);
+    uint8_t* token_fp8_ptr = block_base + pos_in_block * kTokenDataBytes;
+    uint8_t* token_bf16_ptr = token_fp8_ptr + kNopeDim;
+    uint8_t* token_scale_ptr =
+        block_base + static_cast<int64_t>(cache_block_size) * kTokenDataBytes +
+        pos_in_block * kScaleBytesPerToken;
+
+    // Round K to bf16 first, matching the unfused reference path where K is
+    // materialized as bf16 before K quantization.  absmax, clamp, and FP8
+    // quant below all run on these bf16-rounded values.
+#pragma unroll
+    for (int i = 0; i < kElemsPerLane; i++) {
+      elements[i] = Converter::convert(Converter::convert(elements[i]));
+    }
+
+    // Per-quant-block absmax must be computed by ALL 32 lanes (warp-collective
+    // shuffle requires full participation).  RoPE lanes contribute garbage,
+    // but their values are gated out below via `!is_rope_lane`.
+    float local_absmax = 0.0f;
+#pragma unroll
+    for (int i = 0; i < kElemsPerLane; i++) {
+      local_absmax = fmaxf(local_absmax, fabsf(elements[i]));
+    }
+    float const absmax = fmaxf(warp4MaxAbs(local_absmax), 1e-4f);
+    float const exponent = ceilf(log2f(absmax / kFp8Max));
+    float const inv_scale = exp2f(-exponent);
+
+    if (!is_rope_lane) {
+      // ── NoPE lane: UE8M0 FP8 quant ───────────────────────────────────────
+      uint8_t out_bytes[kElemsPerLane];
+#pragma unroll
+      for (int i = 0; i < kElemsPerLane; i++) {
+        float scaled = elements[i] * inv_scale;
+        scaled = fminf(fmaxf(scaled, -kFp8Max), kFp8Max);
+        __nv_fp8_storage_t s =
+            __nv_cvt_float_to_fp8(scaled, __NV_SATFINITE, __NV_E4M3);
+        out_bytes[i] = static_cast<uint8_t>(s);
+      }
+      // One 16-byte STG per lane.
+      *reinterpret_cast<uint4*>(token_fp8_ptr + dim_base) =
+          *reinterpret_cast<uint4 const*>(out_bytes);
+
+      // Lane (4k) of each 4-lane group writes the scale byte for block k<7.
+      if ((laneId & 3) == 0) {
+        int const q_block_idx = laneId >> 2;  // 0..6 for NoPE lanes
+        float encoded = fmaxf(fminf(exponent + 127.0f, 255.0f), 0.0f);
+        token_scale_ptr[q_block_idx] = static_cast<uint8_t>(encoded);
+      }
+      // Lane 0 also writes the padding byte at index 7.
+      if (laneId == 0) {
+        token_scale_ptr[kNumQuantBlocks] = 0;  // pad
+      }
+    } else {
+      // ── RoPE lane: cast back to bf16 and store to cache bf16 tail ────────
+      uint4 out0, out1;
+      typename Converter::packed_hip_type* po0 =
+          reinterpret_cast<typename Converter::packed_hip_type*>(&out0);
+      typename Converter::packed_hip_type* po1 =
+          reinterpret_cast<typename Converter::packed_hip_type*>(&out1);
+#pragma unroll
+      for (int i = 0; i < 4; i++) {
+        po0[i] = Converter::convert(
+            make_float2(elements[2 * i], elements[2 * i + 1]));
+      }
+#pragma unroll
+      for (int i = 0; i < 4; i++) {
+        po1[i] = Converter::convert(
+            make_float2(elements[8 + 2 * i], elements[8 + 2 * i + 1]));
+      }
+      int const rope_local_base = dim_base - kNopeDim;  // in [0, 64)
+      scalar_t_in* bf16_dst =
+          reinterpret_cast<scalar_t_in*>(token_bf16_ptr) + rope_local_base;
+      *reinterpret_cast<uint4*>(bf16_dst) = out0;
+      *reinterpret_cast<uint4*>(bf16_dst + 8) = out1;
+    }
+#if defined(__CUDA_ARCH__) && (__CUDA_ARCH__ >= 900)
+    cudaTriggerProgrammaticLaunchCompletion();
+#endif
+#if (!defined(__CUDA_ARCH__) || __CUDA_ARCH__ < 800) && !defined(USE_ROCM)
+  }
+#endif
+}
+
+// ────────────────────────────────────────────────────────────────────────────
+// Launch wrapper
+// ────────────────────────────────────────────────────────────────────────────
+template <typename scalar_t_in>
+void launchFusedDeepseekV4QNormRopeKVRopeQuantInsert(
+    scalar_t_in* q_inout, scalar_t_in const* kv_in, uint8_t* k_cache,
+    int64_t const* slot_mapping, int64_t const* position_ids,
+    float const* cos_sin_cache, float const eps, int const num_tokens_full,
+    int const num_tokens_insert, int const num_heads_q,
+    int const cache_block_size, int const kv_block_stride,
+    cudaStream_t stream) {
+  constexpr int kBlockSize = 256;
+  constexpr int kWarpsPerBlock = kBlockSize / 32;
+  int64_t const total_warps =
+      static_cast<int64_t>(num_tokens_full) * (num_heads_q + 1);
+  int const grid =
+      static_cast<int>((total_warps + kWarpsPerBlock - 1) / kWarpsPerBlock);
+
+  // PDL: enable programmatic stream serialization whenever the hardware
+  // supports it (SM90+).  On pre-Hopper GPUs the attribute is unavailable,
+  // so leave numAttrs = 0 and launch as a regular kernel.
+  static int const sm_version = getSMVersion();
+  // Host-side guard: the device kernel body is compiled as a no-op for
+  // bf16 on pre-Ampere (sm_70/sm_75) because _typeConvert<BFloat16> is
+  // unavailable there.  Refuse the launch loudly instead of silently
+  // skipping the work.
+  TORCH_CHECK(
+      sm_version >= 80,
+      "fused_deepseek_v4_qnorm_rope_kv_rope_quant_insert requires sm_80+ "
+      "(Ampere or newer); got sm_",
+      sm_version);
+  cudaLaunchConfig_t config;
+  config.gridDim = dim3(grid);
+  config.blockDim = dim3(kBlockSize);
+  config.dynamicSmemBytes = 0;
+  config.stream = stream;
+  cudaLaunchAttribute attrs[1];
+  attrs[0].id = cudaLaunchAttributeProgrammaticStreamSerialization;
+  attrs[0].val.programmaticStreamSerializationAllowed = 1;
+  config.attrs = attrs;
+  config.numAttrs = (sm_version >= 90) ? 1 : 0;
+
+  cudaLaunchKernelEx(
+      &config, fusedDeepseekV4QNormRopeKVRopeQuantInsertKernel<scalar_t_in>,
+      q_inout, kv_in, k_cache, slot_mapping, position_ids, cos_sin_cache, eps,
+      num_tokens_full, num_tokens_insert, num_heads_q, cache_block_size,
+      kv_block_stride);
+}
+
+}  // namespace deepseek_v4_fused_ops
+}  // namespace vllm
+
+// ────────────────────────────────────────────────────────────────────────────
+// Torch op wrapper
+// ────────────────────────────────────────────────────────────────────────────
+void fused_deepseek_v4_qnorm_rope_kv_rope_quant_insert(
+    torch::Tensor& q,                    // [N, H, 512] bf16, in place
+    torch::Tensor const& kv,             // [N, 512] bf16 (read-only)
+    torch::Tensor& k_cache,              // [num_blocks, block_bytes] uint8
+    torch::Tensor const& slot_mapping,   // [N] int64
+    torch::Tensor const& position_ids,   // [N] int64
+    torch::Tensor const& cos_sin_cache,  // [max_pos, rope_dim] bf16
+    double eps, int64_t cache_block_size) {
+  TORCH_CHECK(q.is_cuda() && q.is_contiguous(), "q must be contiguous CUDA");
+  TORCH_CHECK(kv.is_cuda() && kv.is_contiguous(), "kv must be contiguous CUDA");
+  TORCH_CHECK(k_cache.is_cuda(), "k_cache must be CUDA");
+  TORCH_CHECK(slot_mapping.is_cuda() && slot_mapping.dtype() == torch::kInt64,
+              "slot_mapping must be int64 CUDA");
+  TORCH_CHECK(position_ids.is_cuda() && position_ids.dtype() == torch::kInt64,
+              "position_ids must be int64 CUDA");
+  TORCH_CHECK(cos_sin_cache.is_cuda(), "cos_sin_cache must be CUDA");
+  TORCH_CHECK(q.dim() == 3 && q.size(2) == 512, "q shape [N, H, 512]");
+  TORCH_CHECK(kv.dim() == 2 && kv.size(1) == 512, "kv shape [N, 512]");
+  TORCH_CHECK(q.dtype() == kv.dtype(), "q and kv dtype must match");
+  TORCH_CHECK(k_cache.dtype() == torch::kUInt8, "k_cache must be uint8");
+  TORCH_CHECK(cos_sin_cache.dim() == 2 && cos_sin_cache.size(1) == 64,
+              "cos_sin_cache shape [max_pos, 64]");
+  TORCH_CHECK(cos_sin_cache.dtype() == torch::kFloat32,
+              "cos_sin_cache must be float32");
+
+  // With DP padding, slot_mapping can be shorter than q/kv/positions.
+  // Q-norm+RoPE runs on all q.size(0) rows (downstream attention uses them);
+  // KV quant+insert runs only on the first slot_mapping.size(0) rows.
+  int const num_tokens_full = static_cast<int>(q.size(0));
+  int const num_tokens_insert = static_cast<int>(slot_mapping.size(0));
+  TORCH_CHECK(static_cast<int>(kv.size(0)) == num_tokens_full &&
+                  static_cast<int>(position_ids.size(0)) == num_tokens_full,
+              "q/kv/position_ids row counts must match");
+  TORCH_CHECK(num_tokens_insert <= num_tokens_full,
+              "slot_mapping must not exceed q row count");
+  int const num_heads_q = static_cast<int>(q.size(1));
+  int const cache_block_size_i = static_cast<int>(cache_block_size);
+  int const kv_block_stride = static_cast<int>(k_cache.stride(0));
+
+  at::cuda::OptionalCUDAGuard device_guard(device_of(q));
+  auto stream = at::cuda::getCurrentCUDAStream();
+
+  VLLM_DISPATCH_HALF_TYPES(
+      q.scalar_type(), "fused_deepseek_v4_qnorm_rope_kv_insert", [&] {
+        using qkv_scalar_t = scalar_t;
+        vllm::deepseek_v4_fused_ops::
+            launchFusedDeepseekV4QNormRopeKVRopeQuantInsert<qkv_scalar_t>(
+                reinterpret_cast<qkv_scalar_t*>(q.data_ptr()),
+                reinterpret_cast<qkv_scalar_t const*>(kv.data_ptr()),
+                reinterpret_cast<uint8_t*>(k_cache.data_ptr()),
+                reinterpret_cast<int64_t const*>(slot_mapping.data_ptr()),
+                reinterpret_cast<int64_t const*>(position_ids.data_ptr()),
+                cos_sin_cache.data_ptr<float>(), static_cast<float>(eps),
+                num_tokens_full, num_tokens_insert, num_heads_q,
+                cache_block_size_i, kv_block_stride, stream);
+      });
+}

csrc/layernorm_kernels.cu

@@ -77,7 +77,8 @@ __global__ void rms_norm_kernel(
 #pragma unroll
     for (int j = 0; j < VEC_SIZE; j++) {
       float x = static_cast<float>(src1.val[j]);
-      dst.val[j] = ((scalar_t)(x * s_variance)) * src2.val[j];
+      float w = static_cast<float>(src2.val[j]);
+      dst.val[j] = static_cast<scalar_t>(x * s_variance * w);
     }
     v_out[i] = dst;
   }
@@ -134,10 +135,17 @@ fused_add_rms_norm_kernel(
   for (int idx = threadIdx.x; idx < vec_hidden_size; idx += blockDim.x) {
     int id = blockIdx.x * vec_hidden_size + idx;
     int64_t strided_id = blockIdx.x * vec_input_stride + idx;
-    _f16Vec<scalar_t, width> temp = residual_v[id];
-    temp *= s_variance;
-    temp *= weight_v[idx];
-    input_v[strided_id] = temp;
+    _f16Vec<scalar_t, width> res = residual_v[id];
+    _f16Vec<scalar_t, width> w = weight_v[idx];
+    _f16Vec<scalar_t, width> out;
+    using Converter = _typeConvert<scalar_t>;
+#pragma unroll
+    for (int j = 0; j < width; ++j) {
+      float x = Converter::convert(res.data[j]);
+      float wf = Converter::convert(w.data[j]);
+      out.data[j] = Converter::convert(x * s_variance * wf);
+    }
+    input_v[strided_id] = out;
   }
 }
 
@@ -174,8 +182,8 @@ fused_add_rms_norm_kernel(
 
   for (int idx = threadIdx.x; idx < hidden_size; idx += blockDim.x) {
     float x = (float)residual[blockIdx.x * hidden_size + idx];
-    input[blockIdx.x * input_stride + idx] =
-        ((scalar_t)(x * s_variance)) * weight[idx];
+    float w = (float)weight[idx];
+    input[blockIdx.x * input_stride + idx] = (scalar_t)(x * s_variance * w);
   }
 }
 

csrc/layernorm_quant_kernels.cu

@@ -65,9 +65,16 @@ __global__ void rms_norm_static_fp8_quant_kernel(
 #pragma unroll
     for (int j = 0; j < VEC_SIZE; j++) {
       float x = static_cast<float>(src1.val[j]);
-      float const out_norm = ((scalar_t)(x * s_variance)) * src2.val[j];
+      float w = static_cast<float>(src2.val[j]);
+      // Round normalized result through scalar_t to match the precision of the
+      // unfused composite (rms_norm writes scalar_t, then
+      // static_scaled_fp8_quant re-loads it as float before FP8 conversion).
+      // Without this round, the fused path is strictly more accurate and
+      // disagrees with the composite at exact E4M3 quantization tie boundaries.
+      scalar_t out_norm = static_cast<scalar_t>(x * s_variance * w);
       out[blockIdx.x * hidden_size + idx * VEC_SIZE + j] =
-          scaled_fp8_conversion<true, fp8_type>(out_norm, scale_inv);
+          scaled_fp8_conversion<true, fp8_type>(static_cast<float>(out_norm),
+                                                scale_inv);
     }
   }
 }
@@ -127,13 +134,21 @@ fused_add_rms_norm_static_fp8_quant_kernel(
 
   for (int idx = threadIdx.x; idx < vec_hidden_size; idx += blockDim.x) {
     int id = blockIdx.x * vec_hidden_size + idx;
-    _f16Vec<scalar_t, width> temp = residual_v[id];
-    temp *= s_variance;
-    temp *= weight_v[idx];
+    _f16Vec<scalar_t, width> res = residual_v[id];
+    _f16Vec<scalar_t, width> w = weight_v[idx];
+    using Converter = _typeConvert<scalar_t>;
+    using HipT = typename Converter::hip_type;
 #pragma unroll
     for (int i = 0; i < width; ++i) {
-      out[id * width + i] =
-          scaled_fp8_conversion<true, fp8_type>(float(temp.data[i]), scale_inv);
+      float x = Converter::convert(res.data[i]);
+      float wf = Converter::convert(w.data[i]);
+      // See note in rms_norm_static_fp8_quant_kernel: round through scalar_t
+      // to match the unfused composite path at FP8 boundaries. We use the
+      // backend's hip_type for the intermediate since c10::Half/BFloat16 has
+      // ambiguous conversions on CUDA and no implicit conversion on ROCm.
+      HipT out_norm_h = Converter::convert(x * s_variance * wf);
+      out[id * width + i] = scaled_fp8_conversion<true, fp8_type>(
+          Converter::convert(out_norm_h), scale_inv);
     }
   }
 }
@@ -176,9 +191,12 @@ fused_add_rms_norm_static_fp8_quant_kernel(
 
   for (int idx = threadIdx.x; idx < hidden_size; idx += blockDim.x) {
     float x = (float)residual[blockIdx.x * hidden_size + idx];
-    float const out_norm = ((scalar_t)(x * s_variance)) * weight[idx];
-    out[blockIdx.x * hidden_size + idx] =
-        scaled_fp8_conversion<true, fp8_type>(out_norm, scale_inv);
+    float w = (float)weight[idx];
+    // See note in rms_norm_static_fp8_quant_kernel: round through scalar_t
+    // to match the unfused composite path at FP8 boundaries.
+    scalar_t out_norm = static_cast<scalar_t>(x * s_variance * w);
+    out[blockIdx.x * hidden_size + idx] = scaled_fp8_conversion<true, fp8_type>(
+        static_cast<float>(out_norm), scale_inv);
   }
 }
 

csrc/libtorch_stable/ops.h

@@ -134,4 +134,13 @@ void silu_and_mul_nvfp4_quant(torch::stable::Tensor& out,
                               torch::stable::Tensor& input,
                               torch::stable::Tensor& input_global_scale);
 
+void cutlass_mxfp4_group_mm(torch::stable::Tensor& output,
+                            const torch::stable::Tensor& a,
+                            const torch::stable::Tensor& b,
+                            const torch::stable::Tensor& a_blockscale,
+                            const torch::stable::Tensor& b_blockscales,
+                            const torch::stable::Tensor& problem_sizes,
+                            const torch::stable::Tensor& expert_offsets,
+                            const torch::stable::Tensor& sf_offsets);
+
 #endif

csrc/libtorch_stable/quantization/fp4/mxfp4_blockwise_moe_kernel.cu

@@ -0,0 +1,468 @@
+/*
+ * SPDX-License-Identifier: Apache-2.0
+ * SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+ *
+ * MXFP4 x MXFP4 block-scaled grouped GEMM kernel for MoE on SM100.
+ * Uses Cutlass mx_float4_t operands, E8M0 block scales, and 32-element groups.
+ */
+
+#include <torch/csrc/stable/library.h>
+#include <torch/csrc/stable/tensor.h>
+#include "libtorch_stable/torch_utils.h"
+
+#include <cutlass/arch/arch.h>
+
+#include "cutlass_extensions/common.hpp"
+
+#include "cute/tensor.hpp"
+#include "cutlass/tensor_ref.h"
+#include "cutlass/epilogue/collective/default_epilogue.hpp"
+#include "cutlass/epilogue/thread/linear_combination.h"
+#include "cutlass/gemm/dispatch_policy.hpp"
+#include "cutlass/gemm/group_array_problem_shape.hpp"
+#include "cutlass/gemm/collective/collective_builder.hpp"
+#include "cutlass/epilogue/collective/collective_builder.hpp"
+#include "cutlass/gemm/device/gemm_universal_adapter.h"
+#include "cutlass/gemm/kernel/gemm_universal.hpp"
+
+#include "cutlass/util/packed_stride.hpp"
+#include <cassert>
+
+using namespace cute;
+
+// Offset-computation kernel for MXFP4 grouped GEMM (group size 32).
+template <typename ElementAB, typename ElementC, typename ElementSF,
+          typename LayoutSFA, typename LayoutSFB, typename ScaleConfig>
+__global__ void __mxfp4_get_group_gemm_starts(
+    ElementAB** a_offsets, ElementAB** b_offsets, ElementC** out_offsets,
+    ElementSF** a_scales_offsets, ElementSF** b_scales_offsets,
+    LayoutSFA* layout_sfa_base_as_int, LayoutSFB* layout_sfb_base_as_int,
+    ElementAB* a_base_as_int, ElementAB* b_base_as_int,
+    ElementC* out_base_as_int, ElementSF* a_scales_base_as_int,
+    ElementSF* b_scales_base_as_int, const int32_t* expert_offsets,
+    const int32_t* sf_offsets, const int32_t* problem_sizes_as_shapes,
+    int64_t* a_strides, int64_t* b_strides, int64_t* c_strides,
+    const int64_t a_stride_val, const int64_t b_stride_val,
+    const int64_t c_stride_val, const int K, const int N) {
+  int64_t expert_id = threadIdx.x;
+  if (expert_id >= gridDim.x * blockDim.x) {
+    return;
+  }
+  int64_t expert_offset = static_cast<int64_t>(expert_offsets[expert_id]);
+  int64_t sf_offset = static_cast<int64_t>(sf_offsets[expert_id]);
+  int64_t group_size = 32;
+  int64_t m = static_cast<int64_t>(problem_sizes_as_shapes[expert_id * 3]);
+  int64_t n = static_cast<int64_t>(problem_sizes_as_shapes[expert_id * 3 + 1]);
+  int64_t k = static_cast<int64_t>(problem_sizes_as_shapes[expert_id * 3 + 2]);
+  assert((m >= 0 && n == N && k == K && k % 2 == 0) &&
+         "unexpected problem sizes");
+
+  int64_t half_k = static_cast<int64_t>(k / 2);
+  int64_t group_k = static_cast<int64_t>(k / group_size);
+  // Shape of A as uint8/byte = [M, K // 2]
+  a_offsets[expert_id] = a_base_as_int + expert_offset * half_k;
+  // Shape of B as uint8/byte = [E, N, K // 2]
+  b_offsets[expert_id] = b_base_as_int + expert_id * n * half_k;
+  // Shape of C = [M, N]
+  out_offsets[expert_id] = out_base_as_int + expert_offset * n;
+  // Shape of a_scale = [sum(sf_sizes), K // group_size]
+  a_scales_offsets[expert_id] = a_scales_base_as_int + sf_offset * group_k;
+
+  assert((reinterpret_cast<uintptr_t>(a_scales_offsets[expert_id]) % 128) ==
+             0 &&
+         "TMA requires 128-byte alignment");
+
+  // Shape of B scale = [E, N, K // group_size]
+  b_scales_offsets[expert_id] = b_scales_base_as_int + expert_id * n * group_k;
+  assert((reinterpret_cast<uintptr_t>(b_scales_offsets[expert_id]) % 128) ==
+             0 &&
+         "TMA requires 128-byte alignment");
+
+  // Initialize strides
+  a_strides[expert_id] = a_stride_val;
+  b_strides[expert_id] = b_stride_val;
+  c_strides[expert_id] = c_stride_val;
+
+  LayoutSFA* layout_sfa_ptr = layout_sfa_base_as_int + expert_id;
+  LayoutSFB* layout_sfb_ptr = layout_sfb_base_as_int + expert_id;
+
+  *layout_sfa_ptr = ScaleConfig::tile_atom_to_shape_SFA(cute::make_shape(
+      static_cast<int>(m), static_cast<int>(n), static_cast<int>(k), 1));
+  *layout_sfb_ptr = ScaleConfig::tile_atom_to_shape_SFB(cute::make_shape(
+      static_cast<int>(m), static_cast<int>(n), static_cast<int>(k), 1));
+}
+
+#define __CALL_MXFP4_GET_STARTS_KERNEL(ELEMENT_AB_TYPE, SF_TYPE,               \
+                                       TENSOR_C_TYPE, C_TYPE, LayoutSFA,       \
+                                       LayoutSFB, ScaleConfig)                 \
+  else if (out_tensors.scalar_type() == TENSOR_C_TYPE) {                       \
+    __mxfp4_get_group_gemm_starts<ELEMENT_AB_TYPE, C_TYPE, SF_TYPE, LayoutSFA, \
+                                  LayoutSFB, ScaleConfig>                      \
+        <<<1, num_experts, 0, stream>>>(                                       \
+            static_cast<ELEMENT_AB_TYPE**>(a_starts.data_ptr()),               \
+            static_cast<ELEMENT_AB_TYPE**>(b_starts.data_ptr()),               \
+            static_cast<C_TYPE**>(out_starts.data_ptr()),                      \
+            static_cast<SF_TYPE**>(a_scales_starts.data_ptr()),                \
+            static_cast<SF_TYPE**>(b_scales_starts.data_ptr()),                \
+            reinterpret_cast<LayoutSFA*>(layout_sfa.data_ptr()),               \
+            reinterpret_cast<LayoutSFB*>(layout_sfb.data_ptr()),               \
+            static_cast<ELEMENT_AB_TYPE*>(a_tensors.data_ptr()),               \
+            static_cast<ELEMENT_AB_TYPE*>(b_tensors.data_ptr()),               \
+            static_cast<C_TYPE*>(out_tensors.data_ptr()),                      \
+            static_cast<SF_TYPE*>(a_scales.data_ptr()),                        \
+            static_cast<SF_TYPE*>(b_scales.data_ptr()),                        \
+            static_cast<int32_t*>(expert_offsets.data_ptr()),                  \
+            static_cast<int32_t*>(sf_offsets.data_ptr()),                      \
+            static_cast<int32_t*>(problem_sizes.data_ptr()),                   \
+            static_cast<int64_t*>(a_strides.data_ptr()),                       \
+            static_cast<int64_t*>(b_strides.data_ptr()),                       \
+            static_cast<int64_t*>(c_strides.data_ptr()), a_stride_val,         \
+            b_stride_val, c_stride_val, K, N);                                 \
+  }
+
+template <typename LayoutSFA, typename LayoutSFB, typename ScaleConfig>
+void mxfp4_run_get_group_gemm_starts(
+    const torch::stable::Tensor& a_starts,
+    const torch::stable::Tensor& b_starts,
+    const torch::stable::Tensor& out_starts,
+    const torch::stable::Tensor& a_scales_starts,
+    const torch::stable::Tensor& b_scales_starts,
+    const torch::stable::Tensor& layout_sfa,
+    const torch::stable::Tensor& layout_sfb,
+    const torch::stable::Tensor& a_strides,
+    const torch::stable::Tensor& b_strides,
+    const torch::stable::Tensor& c_strides, int64_t a_stride_val,
+    int64_t b_stride_val, int64_t c_stride_val,
+    torch::stable::Tensor const& a_tensors,
+    torch::stable::Tensor const& b_tensors,
+    torch::stable::Tensor const& out_tensors,
+    torch::stable::Tensor const& a_scales,
+    torch::stable::Tensor const& b_scales,
+    torch::stable::Tensor const& expert_offsets,
+    torch::stable::Tensor const& sf_offsets,
+    torch::stable::Tensor const& problem_sizes, int M, int N, int K) {
+  int num_experts = (int)expert_offsets.size(0);
+  auto stream = get_current_cuda_stream(a_tensors.get_device_index());
+
+  STD_TORCH_CHECK(out_tensors.size(1) == N,
+                  "Output tensor shape doesn't match expected shape");
+  STD_TORCH_CHECK(K / 2 == b_tensors.size(2),
+                  "b_tensors(dim = 2) and a_tensors(dim = 1) trailing"
+                  " dimension must match");
+  if (false) {
+  }
+  // MXFP4 uses E8M0 (float_ue8m0_t) scale factors
+  __CALL_MXFP4_GET_STARTS_KERNEL(cutlass::float_e2m1_t, cutlass::float_ue8m0_t,
+                                 torch::headeronly::ScalarType::BFloat16,
+                                 cutlass::bfloat16_t, LayoutSFA, LayoutSFB,
+                                 ScaleConfig)
+  __CALL_MXFP4_GET_STARTS_KERNEL(cutlass::float_e2m1_t, cutlass::float_ue8m0_t,
+                                 torch::headeronly::ScalarType::Half, half,
+                                 LayoutSFA, LayoutSFB, ScaleConfig)
+  else {
+    STD_TORCH_CHECK(false, "Invalid output type (must be float16 or bfloat16)");
+  }
+}
+
+template <typename OutType>
+void run_mxfp4_blockwise_scaled_group_mm_sm100(
+    torch::stable::Tensor& output, const torch::stable::Tensor& a,
+    const torch::stable::Tensor& b, const torch::stable::Tensor& a_blockscale,
+    const torch::stable::Tensor& b_blockscales,
+    const torch::stable::Tensor& problem_sizes,
+    const torch::stable::Tensor& expert_offsets,
+    const torch::stable::Tensor& sf_offsets, int M, int N, int K) {
+  using ProblemShape =
+      cutlass::gemm::GroupProblemShape<Shape<int32_t, int32_t, int32_t>>;
+  using ElementType = cutlass::float_e2m1_t;
+  using ElementSFType = cutlass::float_ue8m0_t;
+  using ElementA = cutlass::mx_float4_t<cutlass::float_e2m1_t>;
+  using ElementB = cutlass::mx_float4_t<cutlass::float_e2m1_t>;
+
+  using ElementC = OutType;
+  using ElementD = ElementC;
+  using ElementAccumulator = float;
+  // Layout definitions
+  using LayoutA = cutlass::layout::RowMajor;
+  using LayoutB = cutlass::layout::ColumnMajor;
+  using LayoutC = cutlass::layout::RowMajor;
+  using LayoutD = LayoutC;
+
+  static constexpr int AlignmentA = 32;
+  static constexpr int AlignmentB = 32;
+  static constexpr int AlignmentC = 128 / cutlass::sizeof_bits<ElementC>::value;
+  static constexpr int AlignmentD = 128 / cutlass::sizeof_bits<ElementD>::value;
+
+  // Architecture definitions
+  using ArchTag = cutlass::arch::Sm100;
+  using EpilogueOperatorClass = cutlass::arch::OpClassTensorOp;
+  using MainloopOperatorClass = cutlass::arch::OpClassBlockScaledTensorOp;
+  using StageCountType = cutlass::gemm::collective::StageCountAuto;
+
+  using ClusterShape = Shape<_1, _1, _1>;
+  struct MMA1SMConfig {
+    using MmaTileShape = Shape<_128, _128, _128>;
+    using KernelSchedule =
+        cutlass::gemm::KernelPtrArrayTmaWarpSpecialized1SmMxf4Sm100;
+    using EpilogueSchedule = cutlass::epilogue::PtrArrayTmaWarpSpecialized1Sm;
+  };
+
+  using CollectiveEpilogue =
+      typename cutlass::epilogue::collective::CollectiveBuilder<
+          ArchTag, EpilogueOperatorClass, typename MMA1SMConfig::MmaTileShape,
+          ClusterShape, Shape<_128, _64>, ElementAccumulator,
+          ElementAccumulator, ElementC, LayoutC*, AlignmentC, ElementD,
+          LayoutC*, AlignmentD,
+          typename MMA1SMConfig::EpilogueSchedule>::CollectiveOp;
+
+  using CollectiveMainloop =
+      typename cutlass::gemm::collective::CollectiveBuilder<
+          ArchTag, MainloopOperatorClass, ElementA, LayoutA*, AlignmentA,
+          ElementB, LayoutB*, AlignmentB, ElementAccumulator,
+          typename MMA1SMConfig::MmaTileShape, ClusterShape,
+          cutlass::gemm::collective::StageCountAutoCarveout<static_cast<int>(
+              sizeof(typename CollectiveEpilogue::SharedStorage))>,
+          typename MMA1SMConfig::KernelSchedule>::CollectiveOp;
+
+  using GemmKernel =
+      cutlass::gemm::kernel::GemmUniversal<ProblemShape, CollectiveMainloop,
+                                           CollectiveEpilogue>;
+
+  using Gemm = cutlass::gemm::device::GemmUniversalAdapter<GemmKernel>;
+  using StrideA = typename Gemm::GemmKernel::InternalStrideA;
+  using StrideB = typename Gemm::GemmKernel::InternalStrideB;
+  using StrideC = typename Gemm::GemmKernel::InternalStrideC;
+  using StrideD = typename Gemm::GemmKernel::InternalStrideD;
+
+  using LayoutSFA =
+      typename Gemm::GemmKernel::CollectiveMainloop::InternalLayoutSFA;
+  using LayoutSFB =
+      typename Gemm::GemmKernel::CollectiveMainloop::InternalLayoutSFB;
+  using ScaleConfig =
+      typename Gemm::GemmKernel::CollectiveMainloop::Sm1xxBlkScaledConfig;
+
+  using UnderlyingProblemShape = ProblemShape::UnderlyingProblemShape;
+  int num_experts = static_cast<int>(expert_offsets.size(0));
+
+  torch::stable::Tensor a_ptrs =
+      torch::stable::empty(num_experts, torch::headeronly::ScalarType::Long,
+                           std::nullopt, a.device());
+  torch::stable::Tensor b_ptrs =
+      torch::stable::empty(num_experts, torch::headeronly::ScalarType::Long,
+                           std::nullopt, a.device());
+  torch::stable::Tensor out_ptrs =
+      torch::stable::empty(num_experts, torch::headeronly::ScalarType::Long,
+                           std::nullopt, a.device());
+  torch::stable::Tensor a_scales_ptrs =
+      torch::stable::empty(num_experts, torch::headeronly::ScalarType::Long,
+                           std::nullopt, a.device());
+  torch::stable::Tensor b_scales_ptrs =
+      torch::stable::empty(num_experts, torch::headeronly::ScalarType::Long,
+                           std::nullopt, a.device());
+  torch::stable::Tensor layout_sfa = torch::stable::empty(
+      {num_experts, 5}, torch::headeronly::ScalarType::Long, std::nullopt,
+      a.device());
+  torch::stable::Tensor layout_sfb = torch::stable::empty(
+      {num_experts, 5}, torch::headeronly::ScalarType::Long, std::nullopt,
+      a.device());
+  torch::stable::Tensor a_strides1 =
+      torch::stable::empty(num_experts, torch::headeronly::ScalarType::Long,
+                           std::nullopt, a.device());
+  torch::stable::Tensor b_strides1 =
+      torch::stable::empty(num_experts, torch::headeronly::ScalarType::Long,
+                           std::nullopt, a.device());
+  torch::stable::Tensor c_strides1 =
+      torch::stable::empty(num_experts, torch::headeronly::ScalarType::Long,
+                           std::nullopt, a.device());
+
+  mxfp4_run_get_group_gemm_starts<LayoutSFA, LayoutSFB, ScaleConfig>(
+      a_ptrs, b_ptrs, out_ptrs, a_scales_ptrs, b_scales_ptrs, layout_sfa,
+      layout_sfb, a_strides1, b_strides1, c_strides1, a.stride(0) * 2,
+      b.stride(1) * 2, output.stride(0), a, b, output, a_blockscale,
+      b_blockscales, expert_offsets, sf_offsets, problem_sizes, M, N, K);
+
+  // Create an instance of the GEMM
+  Gemm gemm_op;
+
+  UnderlyingProblemShape* problem_sizes_as_shapes =
+      static_cast<UnderlyingProblemShape*>(problem_sizes.data_ptr());
+
+  // Set the Scheduler info
+  cutlass::KernelHardwareInfo hw_info;
+  using RasterOrderOptions = typename cutlass::gemm::kernel::detail::
+      PersistentTileSchedulerSm100GroupParams<
+          typename ProblemShape::UnderlyingProblemShape>::RasterOrderOptions;
+  typename Gemm::GemmKernel::TileSchedulerArguments scheduler;
+  scheduler.raster_order = RasterOrderOptions::AlongM;
+  hw_info.device_id = a.get_device_index();
+  static std::unordered_map<int, int> cached_sm_counts;
+  if (cached_sm_counts.find(hw_info.device_id) == cached_sm_counts.end()) {
+    cached_sm_counts[hw_info.device_id] =
+        cutlass::KernelHardwareInfo::query_device_multiprocessor_count(
+            hw_info.device_id);
+  }
+  hw_info.sm_count = min(cached_sm_counts[hw_info.device_id], INT_MAX);
+
+  // Mainloop Arguments
+  typename GemmKernel::MainloopArguments mainloop_args{
+      static_cast<const ElementType**>(a_ptrs.data_ptr()),
+      static_cast<StrideA*>(a_strides1.data_ptr()),
+      static_cast<const ElementType**>(b_ptrs.data_ptr()),
+      static_cast<StrideB*>(b_strides1.data_ptr()),
+      static_cast<const ElementSFType**>(a_scales_ptrs.data_ptr()),
+      reinterpret_cast<LayoutSFA*>(layout_sfa.data_ptr()),
+      static_cast<const ElementSFType**>(b_scales_ptrs.data_ptr()),
+      reinterpret_cast<LayoutSFB*>(layout_sfb.data_ptr())};
+
+  // Epilogue Arguments
+  typename GemmKernel::EpilogueArguments epilogue_args{
+      {},  // epilogue.thread
+      nullptr,
+      static_cast<StrideC*>(c_strides1.data_ptr()),
+      static_cast<ElementD**>(out_ptrs.data_ptr()),
+      static_cast<StrideC*>(c_strides1.data_ptr())};
+  auto& fusion_args = epilogue_args.thread;
+  // Scalar epilogue (CUTLASS grouped GEMM): D = 1 * accum + 0 * C
+  fusion_args.alpha_ptr = nullptr;
+  fusion_args.beta_ptr = nullptr;
+  fusion_args.alpha = 1.0f;
+  fusion_args.alpha_ptr_array = nullptr;
+  fusion_args.dAlpha = {_0{}, _0{}, 0};
+  fusion_args.beta = 0.0f;
+  fusion_args.beta_ptr_array = nullptr;
+  fusion_args.dBeta = {_0{}, _0{}, 0};
+
+  // Gemm Arguments
+  typename GemmKernel::Arguments args{
+      cutlass::gemm::GemmUniversalMode::kGrouped,
+      {num_experts, problem_sizes_as_shapes, nullptr},
+      mainloop_args,
+      epilogue_args,
+      hw_info,
+      scheduler};
+
+  size_t workspace_size = Gemm::get_workspace_size(args);
+  auto workspace =
+      torch::stable::empty(workspace_size, torch::headeronly::ScalarType::Byte,
+                           std::nullopt, a.device());
+  const cudaStream_t stream = get_current_cuda_stream(a.get_device_index());
+
+  auto can_implement_status = gemm_op.can_implement(args);
+  STD_TORCH_CHECK(
+      can_implement_status == cutlass::Status::kSuccess,
+      "Failed to implement MXFP4 GEMM: status=", (int)can_implement_status);
+
+  // Run the GEMM
+  auto status = gemm_op.initialize(args, workspace.data_ptr());
+  STD_TORCH_CHECK(status == cutlass::Status::kSuccess,
+                  "Failed to initialize MXFP4 GEMM: status=", (int)status,
+                  " workspace_size=", workspace_size,
+                  " num_experts=", num_experts, " M=", M, " N=", N, " K=", K);
+
+  status = gemm_op.run(args, workspace.data_ptr(), stream);
+  STD_TORCH_CHECK(status == cutlass::Status::kSuccess,
+                  "Failed to run MXFP4 GEMM");
+}
+
+template <typename OutType>
+void run_mxfp4_blockwise_scaled_group_mm(
+    torch::stable::Tensor& output, const torch::stable::Tensor& a,
+    const torch::stable::Tensor& b, const torch::stable::Tensor& a_blockscale,
+    const torch::stable::Tensor& b_blockscales,
+    const torch::stable::Tensor& problem_sizes,
+    const torch::stable::Tensor& expert_offsets,
+    const torch::stable::Tensor& sf_offsets, int M, int N, int K) {
+  int32_t version_num = get_sm_version_num();
+#if defined ENABLE_NVFP4_SM100 && ENABLE_NVFP4_SM100
+  if (version_num >= 100 && version_num < 120) {
+    run_mxfp4_blockwise_scaled_group_mm_sm100<OutType>(
+        output, a, b, a_blockscale, b_blockscales, problem_sizes,
+        expert_offsets, sf_offsets, M, N, K);
+    return;
+  }
+#endif
+  STD_TORCH_CHECK_NOT_IMPLEMENTED(
+      false,
+      "No compiled cutlass_mxfp4_group_mm kernel for CUDA device capability: ",
+      version_num, ". Required capability: 100");
+}
+
+#if defined ENABLE_NVFP4_SM100 && ENABLE_NVFP4_SM100
+constexpr auto MXFP4_FLOAT4_E2M1X2 = torch::headeronly::ScalarType::Byte;
+// E8M0 scale factors stored as uint8
+constexpr auto MXFP4_SF_DTYPE = torch::headeronly::ScalarType::Byte;
+#endif
+
+#define CHECK_TYPE(x, st, m)             \
+  STD_TORCH_CHECK(x.scalar_type() == st, \
+                  ": Inconsistency of torch::stable::Tensor type:", m)
+#define CHECK_TH_CUDA(x, m) \
+  STD_TORCH_CHECK(x.is_cuda(), m, ": must be a CUDA tensor.")
+#define CHECK_CONTIGUOUS(x, m) \
+  STD_TORCH_CHECK(x.is_contiguous(), m, ": must be contiguous.")
+#define CHECK_INPUT(x, st, m) \
+  CHECK_TH_CUDA(x, m);        \
+  CHECK_CONTIGUOUS(x, m);     \
+  CHECK_TYPE(x, st, m)
+
+void cutlass_mxfp4_group_mm(torch::stable::Tensor& output,
+                            const torch::stable::Tensor& a,
+                            const torch::stable::Tensor& b,
+                            const torch::stable::Tensor& a_blockscale,
+                            const torch::stable::Tensor& b_blockscales,
+                            const torch::stable::Tensor& problem_sizes,
+                            const torch::stable::Tensor& expert_offsets,
+                            const torch::stable::Tensor& sf_offsets) {
+#if defined ENABLE_NVFP4_SM100 && ENABLE_NVFP4_SM100
+  // Input validation
+  CHECK_INPUT(a, MXFP4_FLOAT4_E2M1X2, "a");
+  CHECK_INPUT(b, MXFP4_FLOAT4_E2M1X2, "b");
+  // MXFP4 uses E8M0 scale factors (stored as uint8)
+  CHECK_INPUT(a_blockscale, MXFP4_SF_DTYPE, "a_blockscale");
+  CHECK_INPUT(b_blockscales, MXFP4_SF_DTYPE, "b_blockscales");
+
+  STD_TORCH_CHECK(
+      a_blockscale.dim() == 2,
+      "expected a_blockscale to be of shape [num_experts, rounded_m,"
+      " k // group_size], observed rank: ",
+      a_blockscale.dim())
+  STD_TORCH_CHECK(b_blockscales.dim() == 3,
+                  "expected b_blockscale to be of shape: "
+                  " [num_experts, n, k // group_size], observed rank: ",
+                  b_blockscales.dim())
+  STD_TORCH_CHECK(problem_sizes.dim() == 2,
+                  "problem_sizes must be a 2D tensor");
+  STD_TORCH_CHECK(problem_sizes.size(1) == 3,
+                  "problem_sizes must have the shape (num_experts, 3)");
+  STD_TORCH_CHECK(
+      problem_sizes.size(0) == expert_offsets.size(0),
+      "Number of experts in problem_sizes must match expert_offsets");
+  STD_TORCH_CHECK(
+      problem_sizes.scalar_type() == torch::headeronly::ScalarType::Int,
+      "problem_sizes must be int32.");
+
+  int M = static_cast<int>(a.size(0));
+  int N = static_cast<int>(b.size(1));
+  int E = static_cast<int>(b.size(0));
+  int K = static_cast<int>(2 * b.size(2));
+
+  if (output.scalar_type() == torch::headeronly::ScalarType::BFloat16) {
+    run_mxfp4_blockwise_scaled_group_mm<cutlass::bfloat16_t>(
+        output, a, b, a_blockscale, b_blockscales, problem_sizes,
+        expert_offsets, sf_offsets, M, N, K);
+  } else {
+    run_mxfp4_blockwise_scaled_group_mm<cutlass::half_t>(
+        output, a, b, a_blockscale, b_blockscales, problem_sizes,
+        expert_offsets, sf_offsets, M, N, K);
+  }
+#else
+  STD_TORCH_CHECK_NOT_IMPLEMENTED(
+      false,
+      "No compiled cutlass_mxfp4_group_mm kernel; build vLLM with "
+      "SM100 block-scaled FP4 MoE (ENABLE_NVFP4_SM100) and CUDA 12.8+.");
+#endif
+}
+
+STABLE_TORCH_LIBRARY_IMPL(_C, CUDA, m) {
+  m.impl("cutlass_mxfp4_group_mm", TORCH_BOX(&cutlass_mxfp4_group_mm));
+}

csrc/libtorch_stable/quantization/fp4/mxfp4_experts_quant.cu

@@ -0,0 +1,432 @@
+/*
+ * SPDX-License-Identifier: Apache-2.0
+ * SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+ *
+ * MXFP4 activation quantization kernel for MoE experts.
+ * Quantizes BF16/FP16 activations to MXFP4: E2M1 values with E8M0 block scales
+ * over 32-element groups.
+ *
+ * Uses PACK16 E2M1 conversion helpers (nvfp4_utils.cuh) configured for:
+ *   - Block size 32 (2 threads per SF in PACK16 mode)
+ *   - E8M0 (power-of-two) scale factors
+ *   - SF layout: [numMTiles, numKTiles, 32, 4, 4] where numKTiles=ceil(K/128)
+ */
+
+// MXFP4 requires PACK16 mode (16 elements per thread) so that
+// 2 threads cover 32-element blocks. This requires CUDA >= 12.9.
+// Must be defined before any header that (transitively) includes
+// nvfp4_utils.cuh.
+#define NVFP4_ENABLE_ELTS16 1
+
+#include <cuda.h>
+#include <cuda_runtime_api.h>
+#include <cuda_runtime.h>
+#include <cuda_fp8.h>
+
+#include <torch/csrc/stable/library.h>
+#include <torch/csrc/stable/tensor.h>
+#include "libtorch_stable/torch_utils.h"
+#include "libtorch_stable/dispatch_utils.h"
+#include "cuda_vec_utils.cuh"
+#include "cuda_utils.h"
+
+#include "nvfp4_utils.cuh"
+static_assert(CVT_FP4_ELTS_PER_THREAD == 16,
+              "MXFP4 experts quant requires PACK16 mode (CUDA >= 12.9)");
+
+#include "launch_bounds_utils.h"
+
+namespace vllm {
+
+// MXFP4 block size constants
+static constexpr int MXFP4_SF_VEC_SIZE = 32;
+
+// For PACK16 mode (CVT_FP4_ELTS_PER_THREAD=16): 2 threads per SF
+// For PACK8 mode (CVT_FP4_ELTS_PER_THREAD=8): 4 threads per SF
+static constexpr int MXFP4_NUM_THREADS_PER_SF =
+    MXFP4_SF_VEC_SIZE / CVT_FP4_ELTS_PER_THREAD;
+
+// MXFP4 quantization kernel for experts.
+// Uses 32-element blocks with E8M0 (UE8M0) scale factors.
+// When FUSE_SILU_MUL=true, expects input with gate||up layout and fuses
+// SiLU(gate)*up before quantization.
+template <class Type, bool FUSE_SILU_MUL = false,
+          bool SMALL_NUM_EXPERTS = false>
+__global__ void __launch_bounds__(512, VLLM_BLOCKS_PER_SM(512))
+    mxfp4_cvt_fp16_to_fp4(int32_t numRows, int32_t numCols, Type const* in,
+                          fp4_packed_t* out, uint32_t* SFout,
+                          uint32_t* input_offset_by_experts,
+                          uint32_t* output_scale_offset_by_experts,
+                          int n_experts, bool low_latency) {
+  using PackedVec = PackedVec<Type, CVT_FP4_PACK16>;
+  static_assert(sizeof(PackedVec) == sizeof(Type) * CVT_FP4_ELTS_PER_THREAD,
+                "Vec size is not matched.");
+
+  // MXFP4: numKTiles = ceil(numCols / 128) since block_size=32, 4 SFs/tile
+  int32_t const numKTiles = (numCols + 127) / 128;
+
+  int tid = blockIdx.x * blockDim.x + threadIdx.x;
+  int colsPerRow = numCols / CVT_FP4_ELTS_PER_THREAD;
+  int inColsPerRow = FUSE_SILU_MUL ? colsPerRow * 2 : colsPerRow;
+
+  for (int globalIdx = tid; globalIdx < numRows * colsPerRow;
+       globalIdx += gridDim.x * blockDim.x) {
+    int rowIdx = globalIdx / colsPerRow;
+    int colIdx = globalIdx % colsPerRow;
+
+    int rowIdx_in_expert = 0;
+    int expert_idx = 0;
+
+    if constexpr (SMALL_NUM_EXPERTS) {
+      for (int i = 0; i < n_experts; i++) {
+        uint32_t current_offset = __ldca(&input_offset_by_experts[i]);
+        uint32_t next_offset = __ldca(&input_offset_by_experts[i + 1]);
+        if (rowIdx >= current_offset && rowIdx < next_offset) {
+          rowIdx_in_expert = rowIdx - current_offset;
+          expert_idx = i;
+          break;
+        }
+      }
+    } else {
+      uint32_t local_offsets[17];
+      for (int chunk_start = 0; chunk_start < n_experts; chunk_start += 16) {
+        *reinterpret_cast<int4*>(local_offsets) =
+            __ldca(reinterpret_cast<const int4*>(
+                &input_offset_by_experts[chunk_start]));
+        *reinterpret_cast<int4*>(local_offsets + 4) =
+            __ldca(reinterpret_cast<const int4*>(
+                &input_offset_by_experts[chunk_start + 4]));
+        *reinterpret_cast<int4*>(local_offsets + 8) =
+            __ldca(reinterpret_cast<const int4*>(
+                &input_offset_by_experts[chunk_start + 8]));
+        *reinterpret_cast<int4*>(local_offsets + 12) =
+            __ldca(reinterpret_cast<const int4*>(
+                &input_offset_by_experts[chunk_start + 12]));
+        local_offsets[16] = __ldca(&input_offset_by_experts[chunk_start + 16]);
+
+#pragma unroll
+        for (int i = 0; i < 16; i++) {
+          if (rowIdx >= local_offsets[i] && rowIdx < local_offsets[i + 1]) {
+            rowIdx_in_expert = rowIdx - local_offsets[i];
+            expert_idx = chunk_start + i;
+            break;
+          }
+        }
+      }
+    }
+
+    // Load input and optionally apply fused SiLU+Mul
+    int64_t inOffset = rowIdx * inColsPerRow + colIdx;
+    PackedVec in_vec = reinterpret_cast<PackedVec const*>(in)[inOffset];
+    PackedVec quant_input;
+    if constexpr (FUSE_SILU_MUL) {
+      PackedVec in_vec_up =
+          reinterpret_cast<PackedVec const*>(in)[inOffset + colsPerRow];
+      quant_input = compute_silu_mul(in_vec, in_vec_up);
+    } else {
+      quant_input = in_vec;
+    }
+
+    // In PACK16 mode, each thread outputs 16 E2M1 values = u32x2
+    int64_t outOffset = rowIdx * colsPerRow + colIdx;
+    auto& out_pos = out[outOffset];
+
+    uint32_t* SFout_in_expert =
+        SFout + output_scale_offset_by_experts[expert_idx] * numKTiles;
+
+    // Use MXFP4_NUM_THREADS_PER_SF (2 for PACK16) for 32-element blocks
+    auto sf_out =
+        cvt_quant_to_fp4_get_sf_out_offset<uint32_t, MXFP4_NUM_THREADS_PER_SF>(
+            rowIdx_in_expert, colIdx, numKTiles, SFout_in_expert);
+
+    // Block E8M0 scales only; no extra tensor-level scale in this path
+    constexpr float SFScaleVal = 1.0f;
+    // UE8M0_SF=true for MXFP4 E8M0 scale factors
+    out_pos =
+        cvt_warp_fp16_to_fp4<Type, MXFP4_NUM_THREADS_PER_SF, /*UE8M0_SF=*/true>(
+            quant_input, SFScaleVal, sf_out);
+  }
+}
+
+// Large M_topk variant using shared memory for expert offsets
+template <class Type, bool FUSE_SILU_MUL = false,
+          bool SMALL_NUM_EXPERTS = false>
+__global__ void __launch_bounds__(1024, VLLM_BLOCKS_PER_SM(1024))
+    mxfp4_cvt_fp16_to_fp4(int32_t numRows, int32_t numCols, Type const* in,
+                          fp4_packed_t* out, uint32_t* SFout,
+                          uint32_t* input_offset_by_experts,
+                          uint32_t* output_scale_offset_by_experts,
+                          int n_experts) {
+  using PackedVec = PackedVec<Type, CVT_FP4_PACK16>;
+  static_assert(sizeof(PackedVec) == sizeof(Type) * CVT_FP4_ELTS_PER_THREAD,
+                "Vec size is not matched.");
+
+  // MXFP4: numKTiles = ceil(numCols / 128)
+  int32_t const numKTiles = (numCols + 127) / 128;
+
+  extern __shared__ uint32_t shared_input_offsets[];
+
+  if constexpr (SMALL_NUM_EXPERTS) {
+    for (int i = threadIdx.x; i < n_experts + 1; i += blockDim.x) {
+      shared_input_offsets[i] = input_offset_by_experts[i];
+    }
+  } else {
+    for (int i = threadIdx.x * 4; i < n_experts; i += blockDim.x * 4) {
+      *reinterpret_cast<int4*>(&shared_input_offsets[i]) =
+          *reinterpret_cast<const int4*>(&input_offset_by_experts[i]);
+    }
+    if (threadIdx.x == 0) {
+      shared_input_offsets[n_experts] = input_offset_by_experts[n_experts];
+    }
+  }
+
+  __syncthreads();
+
+  int tid = blockIdx.x * blockDim.x + threadIdx.x;
+  int colsPerRow = numCols / CVT_FP4_ELTS_PER_THREAD;
+  int inColsPerRow = FUSE_SILU_MUL ? colsPerRow * 2 : colsPerRow;
+
+  for (int globalIdx = tid; globalIdx < numRows * colsPerRow;
+       globalIdx += gridDim.x * blockDim.x) {
+    int rowIdx = globalIdx / colsPerRow;
+    int colIdx = globalIdx % colsPerRow;
+
+    int rowIdx_in_expert = 0;
+    int expert_idx = 0;
+
+    // Binary search through experts using shared memory
+    int left = 0, right = n_experts - 1;
+    while (left <= right) {
+      int mid = (left + right) / 2;
+      uint32_t mid_offset = shared_input_offsets[mid];
+      uint32_t next_offset = shared_input_offsets[mid + 1];
+
+      if (rowIdx >= mid_offset && rowIdx < next_offset) {
+        rowIdx_in_expert = rowIdx - mid_offset;
+        expert_idx = mid;
+        break;
+      } else if (rowIdx < mid_offset) {
+        right = mid - 1;
+      } else {
+        left = mid + 1;
+      }
+    }
+
+    int64_t inOffset = rowIdx * inColsPerRow + colIdx;
+    PackedVec in_vec = reinterpret_cast<PackedVec const*>(in)[inOffset];
+    PackedVec quant_input;
+    if constexpr (FUSE_SILU_MUL) {
+      PackedVec in_vec_up =
+          reinterpret_cast<PackedVec const*>(in)[inOffset + colsPerRow];
+      quant_input = compute_silu_mul(in_vec, in_vec_up);
+    } else {
+      quant_input = in_vec;
+    }
+
+    int64_t outOffset = rowIdx * colsPerRow + colIdx;
+    auto& out_pos = out[outOffset];
+
+    // MXFP4 has no global scale - only block-level E8M0 scale factors
+    constexpr float SFScaleVal = 1.0f;
+
+    uint32_t* SFout_in_expert =
+        SFout + output_scale_offset_by_experts[expert_idx] * numKTiles;
+
+    auto sf_out =
+        cvt_quant_to_fp4_get_sf_out_offset<uint32_t, MXFP4_NUM_THREADS_PER_SF>(
+            rowIdx_in_expert, colIdx, numKTiles, SFout_in_expert);
+
+    out_pos =
+        cvt_warp_fp16_to_fp4<Type, MXFP4_NUM_THREADS_PER_SF, /*UE8M0_SF=*/true>(
+            quant_input, SFScaleVal, sf_out);
+  }
+}
+
+template <typename T, bool FUSE_SILU_MUL = false>
+void mxfp4_quant_impl(void* output, void* output_scale, void* input,
+                      void* input_offset_by_experts,
+                      void* output_scale_offset_by_experts, int m_topk, int k,
+                      int n_experts, cudaStream_t stream) {
+  int multiProcessorCount =
+      get_device_attribute(cudaDevAttrMultiProcessorCount, -1);
+
+  int const workSizePerRow = k / ELTS_PER_THREAD;
+  int const totalWorkSize = m_topk * workSizePerRow;
+  dim3 block(std::min(workSizePerRow, 512));
+  int const numBlocksPerSM =
+      vllm_runtime_blocks_per_sm(static_cast<int>(block.x));
+  dim3 grid(std::min(static_cast<int>((totalWorkSize + block.x - 1) / block.x),
+                     multiProcessorCount * numBlocksPerSM));
+  while (grid.x <= multiProcessorCount && block.x > 64) {
+    grid.x *= 2;
+    block.x = (block.x + 1) / 2;
+  }
+
+  int const blockRepeat =
+      (totalWorkSize + block.x * grid.x - 1) / (block.x * grid.x);
+  if (blockRepeat > 1) {
+    size_t shared_mem_size = (n_experts + 1) * sizeof(uint32_t);
+    if (n_experts >= 4) {
+      mxfp4_cvt_fp16_to_fp4<T, FUSE_SILU_MUL, false>
+          <<<grid, block, shared_mem_size, stream>>>(
+              m_topk, k, reinterpret_cast<T*>(input),
+              reinterpret_cast<fp4_packed_t*>(output),
+              reinterpret_cast<uint32_t*>(output_scale),
+              reinterpret_cast<uint32_t*>(input_offset_by_experts),
+              reinterpret_cast<uint32_t*>(output_scale_offset_by_experts),
+              n_experts);
+    } else {
+      mxfp4_cvt_fp16_to_fp4<T, FUSE_SILU_MUL, true>
+          <<<grid, block, shared_mem_size, stream>>>(
+              m_topk, k, reinterpret_cast<T*>(input),
+              reinterpret_cast<fp4_packed_t*>(output),
+              reinterpret_cast<uint32_t*>(output_scale),
+              reinterpret_cast<uint32_t*>(input_offset_by_experts),
+              reinterpret_cast<uint32_t*>(output_scale_offset_by_experts),
+              n_experts);
+    }
+  } else {
+    if (n_experts >= 16) {
+      mxfp4_cvt_fp16_to_fp4<T, FUSE_SILU_MUL, false>
+          <<<grid, block, 0, stream>>>(
+              m_topk, k, reinterpret_cast<T*>(input),
+              reinterpret_cast<fp4_packed_t*>(output),
+              reinterpret_cast<uint32_t*>(output_scale),
+              reinterpret_cast<uint32_t*>(input_offset_by_experts),
+              reinterpret_cast<uint32_t*>(output_scale_offset_by_experts),
+              n_experts, /* bool low_latency */ true);
+    } else {
+      mxfp4_cvt_fp16_to_fp4<T, FUSE_SILU_MUL, true><<<grid, block, 0, stream>>>(
+          m_topk, k, reinterpret_cast<T*>(input),
+          reinterpret_cast<fp4_packed_t*>(output),
+          reinterpret_cast<uint32_t*>(output_scale),
+          reinterpret_cast<uint32_t*>(input_offset_by_experts),
+          reinterpret_cast<uint32_t*>(output_scale_offset_by_experts),
+          n_experts, /* bool low_latency */ true);
+    }
+  }
+}
+
+}  // namespace vllm
+
+/*Quantization entry for mxfp4 experts quantization*/
+#define CHECK_TH_CUDA(x, m) \
+  STD_TORCH_CHECK(x.is_cuda(), m, "must be a CUDA tensor")
+#define CHECK_CONTIGUOUS(x, m) \
+  STD_TORCH_CHECK(x.is_contiguous(), m, "must be contiguous")
+#define CHECK_INPUT(x, m) \
+  CHECK_TH_CUDA(x, m);    \
+  CHECK_CONTIGUOUS(x, m);
+
+constexpr auto HALF = torch::headeronly::ScalarType::Half;
+constexpr auto BF16 = torch::headeronly::ScalarType::BFloat16;
+constexpr auto INT = torch::headeronly::ScalarType::Int;
+constexpr auto UINT8 = torch::headeronly::ScalarType::Byte;
+
+static constexpr int MXFP4_BLOCK_SIZE = 32;
+
+static void validate_mxfp4_experts_quant_inputs(
+    torch::stable::Tensor const& output,
+    torch::stable::Tensor const& output_scale,
+    torch::stable::Tensor const& input,
+    torch::stable::Tensor const& input_offset_by_experts,
+    torch::stable::Tensor const& output_scale_offset_by_experts,
+    int64_t n_experts, int64_t m_topk, int64_t k) {
+  CHECK_INPUT(output, "output");
+  CHECK_INPUT(output_scale, "output_scale");
+  CHECK_INPUT(input, "input");
+  CHECK_INPUT(input_offset_by_experts, "input_offset_by_experts");
+  CHECK_INPUT(output_scale_offset_by_experts, "output_scale_offset_by_experts");
+
+  STD_TORCH_CHECK(output.dim() == 2);
+  STD_TORCH_CHECK(output_scale.dim() == 2);
+  STD_TORCH_CHECK(input.dim() == 2);
+  STD_TORCH_CHECK(input_offset_by_experts.dim() == 1);
+  STD_TORCH_CHECK(output_scale_offset_by_experts.dim() == 1);
+
+  STD_TORCH_CHECK(input.scalar_type() == HALF || input.scalar_type() == BF16);
+  STD_TORCH_CHECK(input_offset_by_experts.scalar_type() == INT);
+  STD_TORCH_CHECK(output_scale_offset_by_experts.scalar_type() == INT);
+  // output is uint8 (two mxfp4 values packed into one uint8)
+  // output_scale is int32 (four E8M0 values packed into one int32)
+  STD_TORCH_CHECK(output.scalar_type() == UINT8);
+  STD_TORCH_CHECK(output_scale.scalar_type() == INT);
+
+  STD_TORCH_CHECK(k % MXFP4_BLOCK_SIZE == 0, "k must be a multiple of 32");
+  STD_TORCH_CHECK(input_offset_by_experts.size(0) == n_experts + 1);
+  STD_TORCH_CHECK(output_scale_offset_by_experts.size(0) == n_experts + 1);
+  STD_TORCH_CHECK(output.size(0) == m_topk);
+  STD_TORCH_CHECK(output.size(1) == k / 2);
+  int scales_k = k / MXFP4_BLOCK_SIZE;
+  // K-dimension scale columns padded to a multiple of 4 for swizzle layout
+  int padded_k = (scales_k + (4 - 1)) / 4 * 4;
+  // 4 = 4 E8M0 values packed into one int32
+  STD_TORCH_CHECK(output_scale.size(1) * 4 == padded_k);
+}
+
+void mxfp4_experts_quant(
+    torch::stable::Tensor& output, torch::stable::Tensor& output_scale,
+    torch::stable::Tensor const& input,
+    torch::stable::Tensor const& input_offset_by_experts,
+    torch::stable::Tensor const& output_scale_offset_by_experts,
+    int64_t n_experts) {
+  auto m_topk = input.size(0);
+  auto k = input.size(1);
+
+  validate_mxfp4_experts_quant_inputs(
+      output, output_scale, input, input_offset_by_experts,
+      output_scale_offset_by_experts, n_experts, m_topk, k);
+
+  const torch::stable::accelerator::DeviceGuard device_guard(
+      input.get_device_index());
+  const cudaStream_t stream = get_current_cuda_stream(input.get_device_index());
+
+  VLLM_STABLE_DISPATCH_HALF_TYPES(
+      input.scalar_type(), "mxfp4_experts_quant_kernel", [&] {
+        using cuda_type = vllm::CUDATypeConverter<scalar_t>::Type;
+        vllm::mxfp4_quant_impl<cuda_type, /*FUSE_SILU_MUL=*/false>(
+            output.data_ptr(), output_scale.data_ptr(), input.data_ptr(),
+            input_offset_by_experts.data_ptr(),
+            output_scale_offset_by_experts.data_ptr(), m_topk, k, n_experts,
+            stream);
+      });
+}
+
+void silu_and_mul_mxfp4_experts_quant(
+    torch::stable::Tensor& output, torch::stable::Tensor& output_scale,
+    torch::stable::Tensor const& input,
+    torch::stable::Tensor const& input_offset_by_experts,
+    torch::stable::Tensor const& output_scale_offset_by_experts,
+    int64_t n_experts) {
+  auto m_topk = input.size(0);
+  auto k_times_2 = input.size(1);
+  STD_TORCH_CHECK(k_times_2 % 2 == 0, "input width must be even (gate || up)");
+  auto k = k_times_2 / 2;
+
+  validate_mxfp4_experts_quant_inputs(
+      output, output_scale, input, input_offset_by_experts,
+      output_scale_offset_by_experts, n_experts, m_topk, k);
+
+  const torch::stable::accelerator::DeviceGuard device_guard(
+      input.get_device_index());
+  const cudaStream_t stream = get_current_cuda_stream(input.get_device_index());
+
+  VLLM_STABLE_DISPATCH_HALF_TYPES(
+      input.scalar_type(), "silu_mul_mxfp4_experts_quant_kernel", [&] {
+        using cuda_type = vllm::CUDATypeConverter<scalar_t>::Type;
+        vllm::mxfp4_quant_impl<cuda_type, /*FUSE_SILU_MUL=*/true>(
+            output.data_ptr(), output_scale.data_ptr(), input.data_ptr(),
+            input_offset_by_experts.data_ptr(),
+            output_scale_offset_by_experts.data_ptr(), m_topk, k, n_experts,
+            stream);
+      });
+}
+
+// Registered here (not torch_bindings.cpp) because VLLM_GPU_FLAGS is applied
+// only under COMPILE_LANGUAGE:CUDA, so ENABLE_NVFP4_SM100 is invisible to
+// .cpp files and cannot gate the registration from there.
+STABLE_TORCH_LIBRARY_IMPL(_C, CUDA, m) {
+  m.impl("mxfp4_experts_quant", TORCH_BOX(&mxfp4_experts_quant));
+  m.impl("silu_and_mul_mxfp4_experts_quant",
+         TORCH_BOX(&silu_and_mul_mxfp4_experts_quant));
+}

csrc/libtorch_stable/quantization/fp4/nvfp4_experts_quant.cu

@@ -277,7 +277,9 @@ void quant_impl(void* output, void* output_scale, void* input,
       (totalWorkSize + block.x * grid.x - 1) / (block.x * grid.x);
   if (blockRepeat > 1) {
     size_t shared_mem_size = (n_experts + 1) * sizeof(uint32_t);
-    if (n_experts >= 4) {
+    // The shared-memory vectorized offset load only handles full 4-expert
+    // chunks. Use the scalar specialization for the remainder cases.
+    if (n_experts >= 4 && n_experts % 4 == 0) {
       cvt_fp16_to_fp4<T, FUSE_SILU_MUL, false, false>
           <<<grid, block, shared_mem_size, stream>>>(
               m_topk, k, reinterpret_cast<T*>(input),
@@ -299,7 +301,9 @@ void quant_impl(void* output, void* output_scale, void* input,
               n_experts);
     }
   } else {
-    if (n_experts >= 16) {
+    // The low-latency vectorized expert lookup only handles full 16-expert
+    // chunks. Fall back to the scalar lookup path for the remainder cases.
+    if (n_experts >= 16 && n_experts % 16 == 0) {
       cvt_fp16_to_fp4<T, FUSE_SILU_MUL, false, false>
           <<<grid, block, 0, stream>>>(
               m_topk, k, reinterpret_cast<T*>(input),

csrc/libtorch_stable/torch_bindings.cpp

@@ -116,6 +116,12 @@ STABLE_TORCH_LIBRARY_FRAGMENT(_C, ops) {
       " Tensor a_blockscale, Tensor b_blockscales, Tensor alphas,"
       " Tensor problem_sizes, Tensor expert_offsets, Tensor sf_offsets) -> ()");
 
+  // cutlass mxfp4 block scaled group GEMM (MXFP4 x MXFP4 MoE)
+  ops.def(
+      "cutlass_mxfp4_group_mm(Tensor! out, Tensor a, Tensor b,"
+      " Tensor a_blockscale, Tensor b_blockscales,"
+      " Tensor problem_sizes, Tensor expert_offsets, Tensor sf_offsets) -> ()");
+
   // Compute NVFP4 block quantized tensor.
   ops.def(
       "scaled_fp4_quant(Tensor input,"
@@ -149,6 +155,19 @@ STABLE_TORCH_LIBRARY_FRAGMENT(_C, ops) {
       "Tensor input, Tensor input_global_scale, Tensor input_offset_by_experts,"
       "Tensor output_scale_offset_by_experts) -> ()");
 
+  // Compute MXFP4 experts quantization (32-element blocks, E8M0 SFs).
+  ops.def(
+      "mxfp4_experts_quant(Tensor! output, Tensor! output_scale,"
+      "Tensor input, Tensor input_offset_by_experts,"
+      "Tensor output_scale_offset_by_experts, int n_experts) -> ()");
+
+  // Fused SiLU+Mul+MXFP4 experts quantization.
+  ops.def(
+      "silu_and_mul_mxfp4_experts_quant(Tensor! output, Tensor! "
+      "output_scale,"
+      "Tensor input, Tensor input_offset_by_experts,"
+      "Tensor output_scale_offset_by_experts, int n_experts) -> ()");
+
   // Fused SiLU+Mul+NVFP4 quantization.
   ops.def(
       "silu_and_mul_nvfp4_quant(Tensor! result, Tensor! result_block_scale, "
@@ -233,9 +252,8 @@ STABLE_TORCH_LIBRARY_IMPL(_C, CUDA, ops) {
   ops.impl("silu_and_mul_scaled_fp4_experts_quant",
            TORCH_BOX(&silu_and_mul_scaled_fp4_experts_quant));
   ops.impl("silu_and_mul_nvfp4_quant", TORCH_BOX(&silu_and_mul_nvfp4_quant));
-
-  // W4A8 ops: impl registrations are in the source files
-  // (w4a8_mm_entry.cu and w4a8_grouped_mm_entry.cu)
+  // mxfp4_experts_quant: registered in mxfp4_experts_quant.cu (SM100 only).
+  // W4A8 ops: registered in w4a8_mm_entry.cu / w4a8_grouped_mm_entry.cu.
 #endif
 }
 

csrc/moe/moe_ops.h

@@ -12,6 +12,15 @@ void topk_sigmoid(torch::Tensor& topk_weights, torch::Tensor& topk_indices,
                   torch::Tensor& gating_output, bool renormalize,
                   std::optional<torch::Tensor> bias);
 
+void topk_softplus_sqrt(torch::Tensor& topk_weights,
+                        torch::Tensor& topk_indices,
+                        torch::Tensor& token_expert_indices,
+                        torch::Tensor& gating_output, bool renormalize,
+                        double routed_scaling_factor,
+                        const c10::optional<torch::Tensor>& correction_bias,
+                        const c10::optional<torch::Tensor>& input_ids,
+                        const c10::optional<torch::Tensor>& tid2eid);
+
 void moe_sum(torch::Tensor& input, torch::Tensor& output);
 
 void moe_align_block_size(torch::Tensor topk_ids, int64_t num_experts,

csrc/moe/topk_softmax_kernels.cu

@@ -126,7 +126,9 @@ __launch_bounds__(TPB) __global__
     {
         const int idx = thread_row_offset + ii;
         const float val = toFloat(input[idx]);
-        const float softmax_val = expf(val - float_max) * normalizing_factor;
+        float softmax_val = expf(val - float_max) * normalizing_factor;
+        // Clamp NaN/Inf to 0 to prevent duplicate expert IDs downstream.
+        if (isnan(softmax_val) || isinf(softmax_val)) softmax_val = 0.f;
         output[idx] = softmax_val;
     }
 }
@@ -147,7 +149,9 @@ __launch_bounds__(TPB) __global__
     {
         const int idx = thread_row_offset + ii;
         const float val = toFloat(input[idx]);
-        const float sigmoid_val = 1.0f / (1.0f + __expf(-val));
+        float sigmoid_val = 1.0f / (1.0f + __expf(-val));
+        // Clamp NaN/Inf to 0 to prevent duplicate expert IDs downstream.
+        if (isnan(sigmoid_val) || isinf(sigmoid_val)) sigmoid_val = 0.f;
         output[idx] = sigmoid_val;
     }
 }
@@ -442,6 +446,19 @@ __launch_bounds__(WARPS_PER_CTA* WARP_SIZE_PARAM) __global__
       }
     }
 
+    // Fix: clamp NaN/Inf values to 0 to prevent duplicate expert IDs.
+    // NaN gating (from degenerate hidden states in CUDA graph padding) causes
+    // softmax to produce all-NaN, which makes the argmax loop always pick
+    // expert 0 for every top-k slot, producing duplicate expert IDs that
+    // crash FlashInfer's three-step MoE sort.
+    // With 0s, the argmax uses index tie-breaking to pick [0,1,2,...,k-1].
+#pragma unroll
+    for (int ii = 0; ii < VPT; ++ii) {
+      if (isnan(row_chunk[ii]) || isinf(row_chunk[ii])) {
+        row_chunk[ii] = 0.f;
+      }
+    }
+
     static constexpr int COLS_PER_GROUP_LDG = ELTS_PER_LDG * THREADS_PER_ROW;
 
     // If bias is not null, use biased value for selection

csrc/moe/topk_softplus_sqrt_kernels.cu

@@ -0,0 +1,715 @@
+/*
+ * Adapted from
+ * https://github.com/NVIDIA/TensorRT-LLM/blob/v0.7.1/cpp/tensorrt_llm/kernels/mixtureOfExperts/moe_kernels.cu
+ * Copyright (c) 2024, The vLLM team.
+ * SPDX-FileCopyrightText: Copyright (c) 1993-2023 NVIDIA CORPORATION &
+ * AFFILIATES. All rights reserved. SPDX-License-Identifier: Apache-2.0
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#include <type_traits>
+#include <torch/all.h>
+#include <ATen/cuda/CUDAContext.h>
+#include <c10/cuda/CUDAGuard.h>
+#include "../cuda_compat.h"
+#include "../cub_helpers.h"
+#ifndef USE_ROCM
+  #include <cuda_bf16.h>
+  #include <cuda_fp16.h>
+#else
+  #include <hip/hip_bf16.h>
+  #include <hip/hip_fp16.h>
+typedef __hip_bfloat16 __nv_bfloat16;
+typedef __hip_bfloat162 __nv_bfloat162;
+#endif
+
+#define MAX(a, b) ((a) > (b) ? (a) : (b))
+#define MIN(a, b) ((a) < (b) ? (a) : (b))
+
+namespace vllm {
+namespace moe {
+
+/// Aligned array type
+template <typename T,
+          /// Number of elements in the array
+          int N,
+          /// Alignment requirement in bytes
+          int Alignment = sizeof(T) * N>
+struct alignas(Alignment) AlignedArray {
+  T data[N];
+};
+
+template <typename T>
+__device__ __forceinline__ float toFloat(T value) {
+  if constexpr (std::is_same_v<T, float>) {
+    return value;
+  } else if constexpr (std::is_same_v<T, __nv_bfloat16>) {
+    return __bfloat162float(value);
+  } else if constexpr (std::is_same_v<T, __half>) {
+    return __half2float(value);
+  }
+}
+
+#define FINAL_MASK 0xffffffff
+template <typename T>
+__inline__ __device__ T warpReduceSum(T val) {
+#pragma unroll
+  for (int mask = 16; mask > 0; mask >>= 1)
+    val += __shfl_xor_sync(FINAL_MASK, val, mask, 32);
+  return val;
+}
+
+// ====================== TopK softplus_sqrt things
+// ===============================
+
+/*
+  A Top-K gating softplus_sqrt written to exploit when the number of experts in
+  the MoE layers are a small power of 2. This allows us to cleanly share the
+  rows among the threads in a single warp and eliminate communication between
+  warps (so no need to use shared mem).
+
+  It fuses the sigmoid, max and argmax into a single kernel.
+
+  Limitations:
+  1) This implementation is optimized for when the number of experts is a small
+  power of 2. Additionally it also supports when number of experts is multiple
+  of 64 which is still faster than the computing sigmoid and topK separately
+  (only tested on CUDA yet). 2) This implementation assumes k is small, but will
+  work for any k.
+*/
+
+template <int VPT, int NUM_EXPERTS, int WARPS_PER_CTA, int BYTES_PER_LDG,
+          int WARP_SIZE_PARAM, bool USE_HASH, typename IndType,
+          typename InputType = float>
+__launch_bounds__(WARPS_PER_CTA* WARP_SIZE_PARAM) __global__
+    void topkGatingSoftplusSqrt(
+        const InputType* input, const bool* finished, float* output,
+        const int num_rows, IndType* indices, int* source_rows, const int k,
+        const int start_expert, const int end_expert, const bool renormalize,
+        double routed_scaling_factor, const float* correction_bias,
+        const IndType* input_ids, const IndType* tid2eid) {
+  static_assert(std::is_same_v<InputType, float> ||
+                    std::is_same_v<InputType, __nv_bfloat16> ||
+                    std::is_same_v<InputType, __half>,
+                "InputType must be float, __nv_bfloat16, or __half");
+
+  // We begin by enforcing compile time assertions and setting up compile time
+  // constants.
+  static_assert(BYTES_PER_LDG == (BYTES_PER_LDG & -BYTES_PER_LDG),
+                "BYTES_PER_LDG must be power of 2");
+  static_assert(BYTES_PER_LDG <= 16, "BYTES_PER_LDG must be leq 16");
+
+  // Number of bytes each thread pulls in per load
+  static constexpr int ELTS_PER_LDG = BYTES_PER_LDG / sizeof(InputType);
+  static constexpr int ELTS_PER_ROW = NUM_EXPERTS;
+  static constexpr int THREADS_PER_ROW = ELTS_PER_ROW / VPT;
+  static constexpr int LDG_PER_THREAD = VPT / ELTS_PER_LDG;
+
+  if constexpr (std::is_same_v<InputType, __nv_bfloat16> ||
+                std::is_same_v<InputType, __half>) {
+    static_assert(ELTS_PER_LDG == 1 || ELTS_PER_LDG % 2 == 0,
+                  "ELTS_PER_LDG must be 1 or even for 16-bit conversion");
+  }
+
+  // Restrictions based on previous section.
+  static_assert(
+      VPT % ELTS_PER_LDG == 0,
+      "The elements per thread must be a multiple of the elements per ldg");
+  static_assert(WARP_SIZE_PARAM % THREADS_PER_ROW == 0,
+                "The threads per row must cleanly divide the threads per warp");
+  static_assert(THREADS_PER_ROW == (THREADS_PER_ROW & -THREADS_PER_ROW),
+                "THREADS_PER_ROW must be power of 2");
+  static_assert(THREADS_PER_ROW <= WARP_SIZE_PARAM,
+                "THREADS_PER_ROW can be at most warp size");
+
+  // We have NUM_EXPERTS elements per row. We specialize for small #experts
+  static constexpr int ELTS_PER_WARP = WARP_SIZE_PARAM * VPT;
+  static constexpr int ROWS_PER_WARP = ELTS_PER_WARP / ELTS_PER_ROW;
+  static constexpr int ROWS_PER_CTA = WARPS_PER_CTA * ROWS_PER_WARP;
+
+  // Restrictions for previous section.
+  static_assert(ELTS_PER_WARP % ELTS_PER_ROW == 0,
+                "The elts per row must cleanly divide the total elt per warp");
+
+  // ===================== From this point, we finally start computing run-time
+  // variables. ========================
+
+  // Compute CTA and warp rows. We pack multiple rows into a single warp, and a
+  // block contains WARPS_PER_CTA warps. This, each block processes a chunk of
+  // rows. We start by computing the start row for each block.
+  const int cta_base_row = blockIdx.x * ROWS_PER_CTA;
+
+  // Now, using the base row per thread block, we compute the base row per warp.
+  const int warp_base_row = cta_base_row + threadIdx.y * ROWS_PER_WARP;
+
+  // The threads in a warp are split into sub-groups that will work on a row.
+  // We compute row offset for each thread sub-group
+  const int thread_row_in_warp = threadIdx.x / THREADS_PER_ROW;
+  const int thread_row = warp_base_row + thread_row_in_warp;
+
+  // Threads with indices out of bounds should early exit here.
+  if (thread_row >= num_rows) {
+    return;
+  }
+  const bool row_is_active = finished ? !finished[thread_row] : true;
+
+  // We finally start setting up the read pointers for each thread. First, each
+  // thread jumps to the start of the row it will read.
+  const InputType* thread_row_ptr = input + thread_row * ELTS_PER_ROW;
+
+  // Now, we compute the group each thread belong to in order to determine the
+  // first column to start loads.
+  const int thread_group_idx = threadIdx.x % THREADS_PER_ROW;
+  const int first_elt_read_by_thread = thread_group_idx * ELTS_PER_LDG;
+  const InputType* thread_read_ptr = thread_row_ptr + first_elt_read_by_thread;
+
+  // Finally, we pull in the data from global mem
+  float row_chunk[VPT];
+
+#if (defined(__CUDA_ARCH__) && (__CUDA_ARCH__ >= 900))
+  asm volatile("griddepcontrol.wait;");
+#endif
+
+  // NOTE(zhuhaoran): dispatch different input types loading, BF16/FP16 convert
+  // to float
+  if constexpr (std::is_same_v<InputType, float>) {
+    using VecType = AlignedArray<float, ELTS_PER_LDG>;
+    VecType* row_chunk_vec_ptr = reinterpret_cast<VecType*>(&row_chunk);
+    const VecType* vec_thread_read_ptr =
+        reinterpret_cast<const VecType*>(thread_read_ptr);
+#pragma unroll
+    for (int ii = 0; ii < LDG_PER_THREAD; ++ii) {
+      row_chunk_vec_ptr[ii] = vec_thread_read_ptr[ii * THREADS_PER_ROW];
+    }
+  } else if constexpr (std::is_same_v<InputType, __nv_bfloat16>) {
+    if constexpr (ELTS_PER_LDG >= 2) {
+      using VecType = AlignedArray<__nv_bfloat16, ELTS_PER_LDG>;
+      float2* row_chunk_f2 = reinterpret_cast<float2*>(row_chunk);
+      const VecType* vec_thread_read_ptr =
+          reinterpret_cast<const VecType*>(thread_read_ptr);
+#pragma unroll
+      for (int ii = 0; ii < LDG_PER_THREAD; ++ii) {
+        VecType vec = vec_thread_read_ptr[ii * THREADS_PER_ROW];
+        int base_idx_f2 = ii * ELTS_PER_LDG / 2;
+#pragma unroll
+        for (int jj = 0; jj < ELTS_PER_LDG / 2; ++jj) {
+          row_chunk_f2[base_idx_f2 + jj] = __bfloat1622float2(
+              *reinterpret_cast<const __nv_bfloat162*>(vec.data + jj * 2));
+        }
+      }
+    } else {  // ELTS_PER_LDG == 1
+#pragma unroll
+      for (int ii = 0; ii < LDG_PER_THREAD; ++ii) {
+        const __nv_bfloat16* scalar_ptr =
+            thread_read_ptr + ii * THREADS_PER_ROW;
+        row_chunk[ii] = __bfloat162float(*scalar_ptr);
+      }
+    }
+  } else if constexpr (std::is_same_v<InputType, __half>) {
+    if constexpr (ELTS_PER_LDG >= 2) {
+      using VecType = AlignedArray<__half, ELTS_PER_LDG>;
+      float2* row_chunk_f2 = reinterpret_cast<float2*>(row_chunk);
+      const VecType* vec_thread_read_ptr =
+          reinterpret_cast<const VecType*>(thread_read_ptr);
+#pragma unroll
+      for (int ii = 0; ii < LDG_PER_THREAD; ++ii) {
+        VecType vec = vec_thread_read_ptr[ii * THREADS_PER_ROW];
+        int base_idx_f2 = ii * ELTS_PER_LDG / 2;
+#pragma unroll
+        for (int jj = 0; jj < ELTS_PER_LDG / 2; ++jj) {
+          row_chunk_f2[base_idx_f2 + jj] = __half22float2(
+              *reinterpret_cast<const __half2*>(vec.data + jj * 2));
+        }
+      }
+    } else {  // ELTS_PER_LDG == 1
+#pragma unroll
+      for (int ii = 0; ii < LDG_PER_THREAD; ++ii) {
+        const __half* scalar_ptr = thread_read_ptr + ii * THREADS_PER_ROW;
+        row_chunk[ii] = __half2float(*scalar_ptr);
+      }
+    }
+  }
+  constexpr float threshold = 20.0f;
+  constexpr float beta = 1.0f;
+
+  // Hash MoE path: indices are predetermined from lookup table
+  if constexpr (USE_HASH) {
+    const IndType token_id = input_ids[thread_row];
+    const IndType* expert_indices_for_token = tid2eid + token_id * k;
+#pragma unroll
+    for (int ii = 0; ii < VPT; ++ii) {
+      float val = row_chunk[ii];
+      float val_b = val * beta;
+      val = (val_b > threshold) ? val : (__logf(1.0f + __expf(val_b))) / beta;
+      row_chunk[ii] = sqrtf(val);
+    }
+    float selected_sum = 0.f;
+#pragma unroll
+    for (int k_idx = 0; k_idx < k; ++k_idx) {
+      const int expert = expert_indices_for_token[k_idx];
+      const int idx = k * thread_row + k_idx;
+      for (int ii = 0; ii < VPT; ++ii) {
+        const int group_id = ii / ELTS_PER_LDG;
+        const int local_id = ii % ELTS_PER_LDG;
+        const int expert_idx = first_elt_read_by_thread +
+                               group_id * THREADS_PER_ROW * ELTS_PER_LDG +
+                               local_id;
+        if (expert == expert_idx) {
+          indices[idx] = expert;
+          selected_sum += row_chunk[ii];
+          break;
+        }
+      }
+    }
+    // Compute per-thread scale (using warp reduction when renormalizing).
+    if (renormalize) {
+      selected_sum = warpReduceSum(selected_sum);
+    }
+    float scale = static_cast<float>(routed_scaling_factor);
+    if (renormalize) {
+      const float denom = selected_sum > 0.f ? selected_sum : 1.f;
+      scale /= denom;
+    }
+
+#pragma unroll
+    for (int k_idx = 0; k_idx < k; ++k_idx) {
+      const int expert = expert_indices_for_token[k_idx];
+      const int idx = k * thread_row + k_idx;
+      for (int ii = 0; ii < VPT; ++ii) {
+        const int group_id = ii / ELTS_PER_LDG;
+        const int local_id = ii % ELTS_PER_LDG;
+        const int expert_idx = first_elt_read_by_thread +
+                               group_id * THREADS_PER_ROW * ELTS_PER_LDG +
+                               local_id;
+        if (expert == expert_idx) {
+          output[idx] = row_chunk[ii] * scale;
+          break;
+        }
+      }
+    }
+#if (defined(__CUDA_ARCH__) && (__CUDA_ARCH__ >= 900))
+    asm volatile("griddepcontrol.launch_dependents;");
+#endif
+    return;
+  }
+
+#pragma unroll
+  for (int ii = 0; ii < VPT; ++ii) {
+    float val = row_chunk[ii];
+    float val_b = val * beta;
+    // Compute softplus: log(1 + exp(val)) with numerical stability
+    // When val > threshold, softplus(x) ≈ x to avoid exp overflow
+    val = (val_b > threshold) ? val : (__logf(1.0f + __expf(val_b))) / beta;
+    val = sqrtf(val);
+    if (correction_bias) {
+      const int group_id = ii / ELTS_PER_LDG;
+      const int local_id = ii % ELTS_PER_LDG;
+      const int expert_idx = first_elt_read_by_thread +
+                             group_id * THREADS_PER_ROW * ELTS_PER_LDG +
+                             local_id;
+      val = val + correction_bias[expert_idx];
+    }
+    row_chunk[ii] = val;
+  }
+
+  // Original TopK path: find top-k experts by score
+  // Now, sigmoid_res contains the sigmoid of the row chunk. Now, I want to find
+  // the topk elements in each row, along with the max index.
+  int start_col = first_elt_read_by_thread;
+  static constexpr int COLS_PER_GROUP_LDG = ELTS_PER_LDG * THREADS_PER_ROW;
+
+  float selected_sum = 0.f;
+  for (int k_idx = 0; k_idx < k; ++k_idx) {
+    // First, each thread does the local argmax
+    float max_val = row_chunk[0];
+    int expert = start_col;
+#pragma unroll
+    for (int ldg = 0, col = start_col; ldg < LDG_PER_THREAD;
+         ++ldg, col += COLS_PER_GROUP_LDG) {
+#pragma unroll
+      for (int ii = 0; ii < ELTS_PER_LDG; ++ii) {
+        float val = row_chunk[ldg * ELTS_PER_LDG + ii];
+
+        // No check on the experts here since columns with the smallest index
+        // are processed first and only updated if > (not >=)
+        if (val > max_val) {
+          max_val = val;
+          expert = col + ii;
+        }
+      }
+    }
+
+// Now, we perform the argmax reduce. We use the butterfly pattern so threads
+// reach consensus about the max. This will be useful for K > 1 so that the
+// threads can agree on "who" had the max value. That thread can then blank out
+// their max with -inf and the warp can run more iterations...
+#pragma unroll
+    for (int mask = THREADS_PER_ROW / 2; mask > 0; mask /= 2) {
+      float other_max =
+          VLLM_SHFL_XOR_SYNC_WIDTH(max_val, mask, THREADS_PER_ROW);
+      int other_expert =
+          VLLM_SHFL_XOR_SYNC_WIDTH(expert, mask, THREADS_PER_ROW);
+
+      // We want lower indices to "win" in every thread so we break ties this
+      // way
+      if (other_max > max_val ||
+          (other_max == max_val && other_expert < expert)) {
+        max_val = other_max;
+        expert = other_expert;
+      }
+    }
+
+    // Write the max for this k iteration to global memory.
+    if (thread_group_idx == 0) {
+      // Add a guard to ignore experts not included by this node
+      const bool node_uses_expert =
+          expert >= start_expert && expert < end_expert;
+      const bool should_process_row = row_is_active && node_uses_expert;
+
+      // The lead thread from each sub-group will write out the final results to
+      // global memory. (This will be a single) thread per row of the
+      // input/output matrices.
+      const int idx = k * thread_row + k_idx;
+      if (correction_bias != nullptr) {
+        max_val -= correction_bias[expert];
+      }
+      output[idx] = max_val;
+      indices[idx] = should_process_row ? (expert - start_expert) : NUM_EXPERTS;
+      source_rows[idx] = k_idx * num_rows + thread_row;
+      if (renormalize) {
+        selected_sum += max_val;
+      }
+    }
+
+    // Finally, we clear the value in the thread with the current max if there
+    // is another iteration to run.
+    if (k_idx + 1 < k) {
+      const int ldg_group_for_expert = expert / COLS_PER_GROUP_LDG;
+      const int thread_to_clear_in_group =
+          (expert / ELTS_PER_LDG) % THREADS_PER_ROW;
+
+      // Only the thread in the group which produced the max will reset the
+      // "winning" value to -inf.
+      if (thread_group_idx == thread_to_clear_in_group) {
+        const int offset_for_expert = expert % ELTS_PER_LDG;
+        // Safe to set to any negative value since row_chunk values must be
+        // between 0 and 1.
+        row_chunk[ldg_group_for_expert * ELTS_PER_LDG + offset_for_expert] =
+            -10000.f;
+      }
+    }
+  }
+
+  // Apply renormalization and routed scaling factor to final weights.
+  if (thread_group_idx == 0) {
+    float scale = static_cast<float>(routed_scaling_factor);
+    if (renormalize) {
+      const float denom = selected_sum > 0.f ? selected_sum : 1.f;
+      scale /= denom;
+    }
+    for (int k_idx = 0; k_idx < k; ++k_idx) {
+      const int idx = k * thread_row + k_idx;
+      output[idx] = output[idx] * scale;
+    }
+  }
+#if (defined(__CUDA_ARCH__) && (__CUDA_ARCH__ >= 900))
+  asm volatile("griddepcontrol.launch_dependents;");
+#endif
+}
+
+namespace detail {
+// Constructs some constants needed to partition the work across threads at
+// compile time.
+template <int EXPERTS, int BYTES_PER_LDG, int WARP_SIZE_PARAM,
+          typename InputType>
+struct TopkConstants {
+  static constexpr int ELTS_PER_LDG = BYTES_PER_LDG / sizeof(InputType);
+  static_assert(EXPERTS / (ELTS_PER_LDG * WARP_SIZE_PARAM) == 0 ||
+                    EXPERTS % (ELTS_PER_LDG * WARP_SIZE_PARAM) == 0,
+                "");
+  static constexpr int VECs_PER_THREAD =
+      MAX(1, EXPERTS / (ELTS_PER_LDG * WARP_SIZE_PARAM));
+  static constexpr int VPT = VECs_PER_THREAD * ELTS_PER_LDG;
+  static constexpr int THREADS_PER_ROW = EXPERTS / VPT;
+  static const int ROWS_PER_WARP = WARP_SIZE_PARAM / THREADS_PER_ROW;
+};
+}  // namespace detail
+
+#define DISPATCH_HASH(use_hash, USE_HASH, ...)                                 \
+  if (use_hash) {                                                              \
+    const bool USE_HASH = true;                                                \
+    static_assert(USE_HASH == true, "USE_HASH must be compile-time constant"); \
+    __VA_ARGS__                                                                \
+  } else {                                                                     \
+    const bool USE_HASH = false;                                               \
+    static_assert(USE_HASH == false,                                           \
+                  "USE_HASH must be compile-time constant");                   \
+    __VA_ARGS__                                                                \
+  }
+
+template <int EXPERTS, int WARPS_PER_TB, int WARP_SIZE_PARAM,
+          int MAX_BYTES_PER_LDG, typename IndType, typename InputType>
+void topkGatingSoftplusSqrtLauncherHelper(
+    const InputType* input, const bool* finished, float* output,
+    IndType* indices, int* source_row, const int num_rows, const int k,
+    const int start_expert, const int end_expert, const bool renormalize,
+    double routed_scaling_factor, const float* correction_bias,
+    const bool use_hash, const IndType* input_ids, const IndType* tid2eid,
+    cudaStream_t stream) {
+  static constexpr int BYTES_PER_LDG =
+      MIN(MAX_BYTES_PER_LDG, sizeof(InputType) * EXPERTS);
+  using Constants =
+      detail::TopkConstants<EXPERTS, BYTES_PER_LDG, WARP_SIZE_PARAM, InputType>;
+  static constexpr int VPT = Constants::VPT;
+  static constexpr int ROWS_PER_WARP = Constants::ROWS_PER_WARP;
+  const int num_warps = (num_rows + ROWS_PER_WARP - 1) / ROWS_PER_WARP;
+  const int num_blocks = (num_warps + WARPS_PER_TB - 1) / WARPS_PER_TB;
+  dim3 block_dim(WARP_SIZE_PARAM, WARPS_PER_TB);
+  DISPATCH_HASH(use_hash, USE_HASH, {
+    auto* kernel =
+        &topkGatingSoftplusSqrt<VPT, EXPERTS, WARPS_PER_TB, BYTES_PER_LDG,
+                                WARP_SIZE_PARAM, USE_HASH, IndType, InputType>;
+#ifndef USE_ROCM
+    cudaLaunchConfig_t config = {};
+    config.gridDim = num_blocks;
+    config.blockDim = block_dim;
+    config.dynamicSmemBytes = 0;
+    config.stream = stream;
+    cudaLaunchAttribute attrs[1];
+    attrs[0].id = cudaLaunchAttributeProgrammaticStreamSerialization;
+    attrs[0].val.programmaticStreamSerializationAllowed = 1;
+    config.numAttrs = 1;
+    config.attrs = attrs;
+    cudaLaunchKernelEx(&config, kernel, input, finished, output, num_rows,
+                       indices, source_row, k, start_expert, end_expert,
+                       renormalize, routed_scaling_factor, correction_bias,
+                       input_ids, tid2eid);
+#else
+    kernel<<<num_blocks, block_dim, 0, stream>>>(
+        input, finished, output, num_rows, indices, source_row, k, start_expert,
+        end_expert, renormalize, routed_scaling_factor, correction_bias,
+        input_ids, tid2eid);
+#endif
+  })
+}
+
+#ifndef USE_ROCM
+  #define LAUNCH_SOFTPLUS_SQRT(NUM_EXPERTS, WARPS_PER_TB, MAX_BYTES)           \
+    static_assert(WARP_SIZE == 32,                                             \
+                  "Unsupported warp size. Only 32 is supported for CUDA");     \
+    topkGatingSoftplusSqrtLauncherHelper<NUM_EXPERTS, WARPS_PER_TB, WARP_SIZE, \
+                                         MAX_BYTES>(                           \
+        gating_output, nullptr, topk_weights, topk_indices,                    \
+        token_expert_indices, num_tokens, topk, 0, num_experts, renormalize,   \
+        routed_scaling_factor, correction_bias, use_hash, input_ids, tid2eid,  \
+        stream);
+#else
+  #define LAUNCH_SOFTPLUS_SQRT(NUM_EXPERTS, WARPS_PER_TB, MAX_BYTES)           \
+    if (WARP_SIZE == 64) {                                                     \
+      topkGatingSoftplusSqrtLauncherHelper<NUM_EXPERTS, WARPS_PER_TB, 64,      \
+                                           MAX_BYTES>(                         \
+          gating_output, nullptr, topk_weights, topk_indices,                  \
+          token_expert_indices, num_tokens, topk, 0, num_experts, renormalize, \
+          routed_scaling_factor, correction_bias, use_hash, input_ids,         \
+          tid2eid, stream);                                                    \
+    } else if (WARP_SIZE == 32) {                                              \
+      topkGatingSoftplusSqrtLauncherHelper<NUM_EXPERTS, WARPS_PER_TB, 32,      \
+                                           MAX_BYTES>(                         \
+          gating_output, nullptr, topk_weights, topk_indices,                  \
+          token_expert_indices, num_tokens, topk, 0, num_experts, renormalize, \
+          routed_scaling_factor, correction_bias, use_hash, input_ids,         \
+          tid2eid, stream);                                                    \
+    } else {                                                                   \
+      assert(false &&                                                          \
+             "Unsupported warp size. Only 32 and 64 are supported for ROCm");  \
+    }
+#endif
+
+template <typename IndType, typename InputType>
+void topkGatingSoftplusSqrtKernelLauncher(
+    const InputType* gating_output, float* topk_weights, IndType* topk_indices,
+    int* token_expert_indices, const int num_tokens, const int num_experts,
+    const int topk, const bool renormalize, double routed_scaling_factor,
+    const float* correction_bias, const bool use_hash, const IndType* input_ids,
+    const IndType* tid2eid, cudaStream_t stream) {
+  static constexpr int WARPS_PER_TB = 4;
+  static constexpr int BYTES_PER_LDG_POWER_OF_2 = 16;
+#ifndef USE_ROCM
+  // for bfloat16 dtype, we need 4 bytes loading to make sure num_experts
+  // elements can be loaded by a warp
+  static constexpr int BYTES_PER_LDG_MULTIPLE_64 =
+      (std::is_same_v<InputType, __nv_bfloat16> ||
+       std::is_same_v<InputType, __half>)
+          ? 4
+          : 8;
+#endif
+  switch (num_experts) {
+    case 1:
+      LAUNCH_SOFTPLUS_SQRT(1, WARPS_PER_TB, BYTES_PER_LDG_POWER_OF_2);
+      break;
+    case 2:
+      LAUNCH_SOFTPLUS_SQRT(2, WARPS_PER_TB, BYTES_PER_LDG_POWER_OF_2);
+      break;
+    case 4:
+      LAUNCH_SOFTPLUS_SQRT(4, WARPS_PER_TB, BYTES_PER_LDG_POWER_OF_2);
+      break;
+    case 8:
+      LAUNCH_SOFTPLUS_SQRT(8, WARPS_PER_TB, BYTES_PER_LDG_POWER_OF_2);
+      break;
+    case 16:
+      LAUNCH_SOFTPLUS_SQRT(16, WARPS_PER_TB, BYTES_PER_LDG_POWER_OF_2);
+      break;
+    case 32:
+      LAUNCH_SOFTPLUS_SQRT(32, WARPS_PER_TB, BYTES_PER_LDG_POWER_OF_2);
+      break;
+    case 64:
+      LAUNCH_SOFTPLUS_SQRT(64, WARPS_PER_TB, BYTES_PER_LDG_POWER_OF_2);
+      break;
+    case 128:
+      LAUNCH_SOFTPLUS_SQRT(128, WARPS_PER_TB, BYTES_PER_LDG_POWER_OF_2);
+      break;
+    case 256:
+      LAUNCH_SOFTPLUS_SQRT(256, WARPS_PER_TB, BYTES_PER_LDG_POWER_OF_2);
+      break;
+    case 512:
+      LAUNCH_SOFTPLUS_SQRT(512, WARPS_PER_TB, BYTES_PER_LDG_POWER_OF_2);
+      break;
+      // (CUDA only) support multiples of 64 when num_experts is not power of 2.
+      // ROCm uses WARP_SIZE 64 so 8 bytes loading won't fit for some of
+      // num_experts, alternatively we can test 4 bytes loading and enable it in
+      // future.
+#ifndef USE_ROCM
+    case 192:
+      LAUNCH_SOFTPLUS_SQRT(192, WARPS_PER_TB, BYTES_PER_LDG_MULTIPLE_64);
+      break;
+    case 320:
+      LAUNCH_SOFTPLUS_SQRT(320, WARPS_PER_TB, BYTES_PER_LDG_MULTIPLE_64);
+      break;
+    case 384:
+      LAUNCH_SOFTPLUS_SQRT(384, WARPS_PER_TB, BYTES_PER_LDG_MULTIPLE_64);
+      break;
+    case 448:
+      LAUNCH_SOFTPLUS_SQRT(448, WARPS_PER_TB, BYTES_PER_LDG_MULTIPLE_64);
+      break;
+    case 576:
+      LAUNCH_SOFTPLUS_SQRT(576, WARPS_PER_TB, BYTES_PER_LDG_MULTIPLE_64);
+      break;
+#endif
+    default: {
+      TORCH_CHECK(false, "Unsupported expert number: ", num_experts);
+    }
+  }
+}
+
+}  // namespace moe
+}  // namespace vllm
+
+template <typename ComputeType>
+void dispatch_topk_softplus_sqrt_launch(
+    const ComputeType* gating_output, torch::Tensor& topk_weights,
+    torch::Tensor& topk_indices, torch::Tensor& token_expert_indices,
+    int num_tokens, int num_experts, int topk, bool renormalize,
+    double routed_scaling_factor,
+    const c10::optional<torch::Tensor>& correction_bias,
+    const c10::optional<torch::Tensor>& input_ids,
+    const c10::optional<torch::Tensor>& tid2eid, cudaStream_t stream) {
+  const float* bias_ptr = nullptr;
+  if (correction_bias.has_value()) {
+    bias_ptr = correction_bias.value().data_ptr<float>();
+  }
+  bool use_hash = false;
+  if (tid2eid.has_value()) {
+    TORCH_CHECK(input_ids.has_value(), "input_ids is required for hash MoE");
+    use_hash = true;
+  }
+  if (topk_indices.scalar_type() == at::ScalarType::Int) {
+    const int* input_ids_ptr = nullptr;
+    const int* tid2eid_ptr = nullptr;
+    if (tid2eid.has_value()) {
+      input_ids_ptr = input_ids.value().data_ptr<int>();
+      tid2eid_ptr = tid2eid.value().data_ptr<int>();
+    }
+
+    vllm::moe::topkGatingSoftplusSqrtKernelLauncher<int, ComputeType>(
+        gating_output, topk_weights.data_ptr<float>(),
+        topk_indices.data_ptr<int>(), token_expert_indices.data_ptr<int>(),
+        num_tokens, num_experts, topk, renormalize, routed_scaling_factor,
+        bias_ptr, use_hash, input_ids_ptr, tid2eid_ptr, stream);
+  } else if (topk_indices.scalar_type() == at::ScalarType::UInt32) {
+    const uint32_t* input_ids_ptr = nullptr;
+    const uint32_t* tid2eid_ptr = nullptr;
+    if (tid2eid.has_value()) {
+      input_ids_ptr = input_ids.value().data_ptr<uint32_t>();
+      tid2eid_ptr = tid2eid.value().data_ptr<uint32_t>();
+    }
+    vllm::moe::topkGatingSoftplusSqrtKernelLauncher<uint32_t, ComputeType>(
+        gating_output, topk_weights.data_ptr<float>(),
+        topk_indices.data_ptr<uint32_t>(), token_expert_indices.data_ptr<int>(),
+        num_tokens, num_experts, topk, renormalize, routed_scaling_factor,
+        bias_ptr, use_hash, input_ids_ptr, tid2eid_ptr, stream);
+  } else {
+    TORCH_CHECK(topk_indices.scalar_type() == at::ScalarType::Long);
+
+    const int64_t* input_ids_ptr = nullptr;
+    const int64_t* tid2eid_ptr = nullptr;
+    if (tid2eid.has_value()) {
+      input_ids_ptr = input_ids.value().data_ptr<int64_t>();
+      tid2eid_ptr = tid2eid.value().data_ptr<int64_t>();
+    }
+
+    vllm::moe::topkGatingSoftplusSqrtKernelLauncher<int64_t, ComputeType>(
+        gating_output, topk_weights.data_ptr<float>(),
+        topk_indices.data_ptr<int64_t>(), token_expert_indices.data_ptr<int>(),
+        num_tokens, num_experts, topk, renormalize, routed_scaling_factor,
+        bias_ptr, use_hash, input_ids_ptr, tid2eid_ptr, stream);
+  }
+}
+
+void topk_softplus_sqrt(
+    torch::Tensor& topk_weights,          // [num_tokens, topk]
+    torch::Tensor& topk_indices,          // [num_tokens, topk]
+    torch::Tensor& token_expert_indices,  // [num_tokens, topk]
+    torch::Tensor& gating_output,         // [num_tokens, num_experts]
+    bool renormalize, double routed_scaling_factor,
+    const c10::optional<torch::Tensor>& correction_bias,
+    const c10::optional<torch::Tensor>& input_ids,
+    const c10::optional<torch::Tensor>& tid2eid) {
+  const int num_experts = gating_output.size(-1);
+  const auto num_tokens = gating_output.numel() / num_experts;
+  const int topk = topk_weights.size(-1);
+  const at::cuda::OptionalCUDAGuard device_guard(device_of(gating_output));
+  const cudaStream_t stream = at::cuda::getCurrentCUDAStream();
+
+  if (gating_output.scalar_type() == at::ScalarType::Float) {
+    dispatch_topk_softplus_sqrt_launch<float>(
+        gating_output.data_ptr<float>(), topk_weights, topk_indices,
+        token_expert_indices, num_tokens, num_experts, topk, renormalize,
+        routed_scaling_factor, correction_bias, input_ids, tid2eid, stream);
+  } else if (gating_output.scalar_type() == at::ScalarType::Half) {
+    dispatch_topk_softplus_sqrt_launch<__half>(
+        reinterpret_cast<const __half*>(gating_output.data_ptr<at::Half>()),
+        topk_weights, topk_indices, token_expert_indices, num_tokens,
+        num_experts, topk, renormalize, routed_scaling_factor, correction_bias,
+        input_ids, tid2eid, stream);
+  } else if (gating_output.scalar_type() == at::ScalarType::BFloat16) {
+    dispatch_topk_softplus_sqrt_launch<__nv_bfloat16>(
+        reinterpret_cast<const __nv_bfloat16*>(
+            gating_output.data_ptr<at::BFloat16>()),
+        topk_weights, topk_indices, token_expert_indices, num_tokens,
+        num_experts, topk, renormalize, routed_scaling_factor, correction_bias,
+        input_ids, tid2eid, stream);
+  } else {
+    TORCH_CHECK(false, "Unsupported gating_output data type: ",
+                gating_output.scalar_type());
+  }
+}

csrc/moe/torch_bindings.cpp

@@ -16,6 +16,14 @@ TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, m) {
       "bias) -> ()");
   m.impl("topk_sigmoid", torch::kCUDA, &topk_sigmoid);
 
+#ifndef USE_ROCM
+  m.def(
+      "topk_softplus_sqrt(Tensor! topk_weights, Tensor! topk_indices, Tensor! "
+      "token_expert_indices, Tensor gating_output, bool renormalize, float "
+      "routed_scaling_factor, Tensor? "
+      "bias, Tensor? input_ids, Tensor? tid2eid) -> ()");
+  m.impl("topk_softplus_sqrt", torch::kCUDA, &topk_softplus_sqrt);
+#endif
   // Calculate the result of moe by summing up the partial results
   // from all selected experts.
   m.def("moe_sum(Tensor input, Tensor! output) -> ()");

csrc/ops.h

@@ -100,6 +100,11 @@ void fused_qk_norm_rope(torch::Tensor& qkv, int64_t num_heads_q,
                         bool is_neox, torch::Tensor& position_ids,
                         int64_t forced_token_heads_per_warp);
 
+void fused_deepseek_v4_qnorm_rope_kv_rope_quant_insert(
+    torch::Tensor& q, torch::Tensor const& kv, torch::Tensor& k_cache,
+    torch::Tensor const& slot_mapping, torch::Tensor const& position_ids,
+    torch::Tensor const& cos_sin_cache, double eps, int64_t cache_block_size);
+
 void apply_repetition_penalties_(torch::Tensor& logits,
                                  const torch::Tensor& prompt_mask,
                                  const torch::Tensor& output_mask,
@@ -153,10 +158,13 @@ void silu_and_mul_per_block_quant(torch::Tensor& out,
 
 void rotary_embedding(torch::Tensor& positions, torch::Tensor& query,
                       std::optional<torch::Tensor> key, int64_t head_size,
-                      torch::Tensor& cos_sin_cache, bool is_neox);
+                      torch::Tensor& cos_sin_cache, bool is_neox,
+                      int64_t rope_dim_offset, bool inverse);
 
 void silu_and_mul(torch::Tensor& out, torch::Tensor& input);
 
+void silu_and_mul_clamp(torch::Tensor& out, torch::Tensor& input, double limit);
+
 void silu_and_mul_quant(torch::Tensor& out, torch::Tensor& input,
                         torch::Tensor& scale);
 

csrc/persistent_topk.cuh

@@ -18,7 +18,6 @@ namespace persistent {
 // Constants
 // ============================================================================
 
-constexpr int TopK = 2048;
 constexpr int kThreadsPerBlock = 1024;
 constexpr int RADIX = 256;
 
@@ -128,11 +127,12 @@ struct RadixRowState {
 
 struct PersistentTopKParams {
   const float* __restrict__ input;  // [num_rows, stride]
-  int32_t* __restrict__ output;     // [num_rows, TopK]
+  int32_t* __restrict__ output;     // [num_rows, top_k]
   int32_t* __restrict__ lengths;    // [num_rows]
   RadixRowState* row_states;        // large path: per-group state
   uint32_t num_rows;
   uint32_t stride;
+  uint32_t top_k;           // actual k value for output stride
   uint32_t chunk_size;      // large path: elements per CTA
   uint32_t ctas_per_group;  // 1=medium, >1=large
   uint32_t max_seq_len;     // max seq_len across all rows (for early CTA exit)
@@ -154,6 +154,7 @@ __device__ __forceinline__ uint32_t decode_bin(float x) {
   return key >> 5;
 }
 
+template <int TopK>
 __device__ __noinline__ void histogram_2048_topk(
     const float* __restrict__ logits, int32_t* __restrict__ output_indices,
     int32_t seq_len) {
@@ -418,6 +419,7 @@ __device__ __noinline__ void histogram_2048_topk(
 // by: DarkSharpness
 // which at the same time is an optimized topk kernel copied from tilelang
 // kernel
+template <int TopK>
 __device__ __noinline__ void histogram_256_topk(
     const float* __restrict__ logits, int* __restrict__ output_indices,
     int logits_offset, int seq_len) {
@@ -649,7 +651,7 @@ __device__ __forceinline__ void wait_ge(int* ptr, int target_val,
 // Adapted from https://github.com/flashinfer-ai/flashinfer/pull/2215
 // ============================================================================
 
-template <uint32_t VEC_SIZE>
+template <int TopK, uint32_t VEC_SIZE>
 __device__ void radix_topk(const float* __restrict__ row_input,
                            int32_t* __restrict__ row_output, uint32_t seq_len,
                            uint32_t my_chunk_start, uint32_t chunk_size,
@@ -857,7 +859,7 @@ __device__ void radix_topk(const float* __restrict__ row_input,
 // see filtered_topk.cuh)
 // ============================================================================
 
-template <uint32_t VEC_SIZE = 1>
+template <int TopK = 2048, uint32_t VEC_SIZE = 1>
 __global__ void __launch_bounds__(kThreadsPerBlock, 2)
     persistent_topk_kernel(PersistentTopKParams params) {
   const uint32_t tx = threadIdx.x;
@@ -915,7 +917,7 @@ __global__ void __launch_bounds__(kThreadsPerBlock, 2)
     if (row_idx >= params.num_rows) break;
 
     const uint32_t seq_len = params.lengths[row_idx];
-    int32_t* row_output = params.output + row_idx * TopK;
+    int32_t* row_output = params.output + row_idx * params.top_k;
     const float* row_input = params.input + row_idx * params.stride;
 
     if (seq_len <= RADIX_THRESHOLD) {
@@ -927,19 +929,19 @@ __global__ void __launch_bounds__(kThreadsPerBlock, 2)
             row_output[i] = (i < seq_len) ? static_cast<int32_t>(i) : -1;
           }
         } else if (seq_len <= static_cast<uint32_t>(HIST2048_THRESHOLD)) {
-          histogram_2048_topk(row_input, row_output, seq_len);
+          histogram_2048_topk<TopK>(row_input, row_output, seq_len);
         } else {
-          histogram_256_topk(row_input, row_output, 0, seq_len);
+          histogram_256_topk<TopK>(row_input, row_output, 0, seq_len);
         }
       }
       continue;
     }
 
     const uint32_t my_chunk_start = cta_in_group * chunk_size;
-    radix_topk<VEC_SIZE>(row_input, row_output, seq_len, my_chunk_start,
-                         chunk_size, local_histogram, suffix_sum,
-                         shared_scalars, shared_ordered, state, cta_in_group,
-                         ctas_per_group, barrier_phase, iter, tx);
+    radix_topk<TopK, VEC_SIZE>(
+        row_input, row_output, seq_len, my_chunk_start, chunk_size,
+        local_histogram, suffix_sum, shared_scalars, shared_ordered, state,
+        cta_in_group, ctas_per_group, barrier_phase, iter, tx);
   }
 }
 
@@ -1011,7 +1013,6 @@ struct FilteredTopKTraits<float> {
   }
 };
 
-constexpr uint32_t FILTERED_TOPK_MAX_K = 2048;
 constexpr uint32_t FILTERED_TOPK_BLOCK_THREADS = 1024;
 constexpr uint32_t FILTERED_TOPK_SMEM_INPUT_SIZE =
     16 * 1024;  // 16K indices per buffer
@@ -1025,7 +1026,7 @@ constexpr size_t FILTERED_TOPK_SMEM_DYNAMIC =
  * \tparam IdType Index type (int32_t)
  * \tparam VEC_SIZE Vector size for input loads (1, 2, 4, or 8)
  */
-template <typename DType, typename IdType, int VEC_SIZE>
+template <typename DType, typename IdType, int VEC_SIZE, uint32_t MAX_K = 2048>
 __global__ void __launch_bounds__(FILTERED_TOPK_BLOCK_THREADS)
     FilteredTopKUnifiedKernel(const DType* __restrict__ input,
                               IdType* __restrict__ output,
@@ -1059,7 +1060,7 @@ __global__ void __launch_bounds__(FILTERED_TOPK_BLOCK_THREADS)
   alignas(128) __shared__ int s_counter;
   alignas(128) __shared__ int s_threshold_bin_id;
   alignas(128) __shared__ int s_num_input[2];
-  alignas(128) __shared__ int s_indices[FILTERED_TOPK_MAX_K];
+  alignas(128) __shared__ int s_indices[MAX_K];
 
   auto& s_histogram = s_histogram_buf[0];
 
@@ -1280,7 +1281,7 @@ constexpr int ComputeFilteredTopKVecSize(uint32_t max_len) {
   return static_cast<int>(g);
 }
 
-template <typename DType, typename IdType>
+template <typename DType, typename IdType, uint32_t MAX_K = 2048>
 cudaError_t FilteredTopKRaggedTransform(DType* input, IdType* output_indices,
                                         IdType* lengths, uint32_t num_rows,
                                         uint32_t top_k_val, uint32_t max_len,
@@ -1297,7 +1298,7 @@ cudaError_t FilteredTopKRaggedTransform(DType* input, IdType* output_indices,
 
 #define DISPATCH_VEC_SIZE(VS)                                               \
   if (vec_size == VS) {                                                     \
-    auto kernel = FilteredTopKUnifiedKernel<DType, IdType, VS>;             \
+    auto kernel = FilteredTopKUnifiedKernel<DType, IdType, VS, MAX_K>;      \
     FLASHINFER_CUDA_CALL(cudaFuncSetAttribute(                              \
         kernel, cudaFuncAttributeMaxDynamicSharedMemorySize, smem_size));   \
     FLASHINFER_CUDA_CALL(cudaLaunchKernel((void*)kernel, grid, block, args, \

csrc/pos_encoding_kernels.cu

@@ -9,28 +9,29 @@ namespace vllm {
 
 template <typename scalar_t, bool IS_NEOX>
 inline __device__ void apply_token_rotary_embedding(
-    scalar_t* __restrict__ arr, const scalar_t* __restrict__ cos_ptr,
-    const scalar_t* __restrict__ sin_ptr, int rot_offset, int embed_dim) {
+    scalar_t* __restrict__ arr, const float* __restrict__ cos_ptr,
+    const float* __restrict__ sin_ptr, int rot_offset, int embed_dim,
+    const bool inverse) {
   int x_index, y_index;
-  scalar_t cos, sin;
+  float cos_f, sin_f;
   if (IS_NEOX) {
-    // GPT-NeoX style rotary embedding.
     x_index = rot_offset;
     y_index = embed_dim + rot_offset;
-    cos = VLLM_LDG(cos_ptr + x_index);
-    sin = VLLM_LDG(sin_ptr + x_index);
+    cos_f = VLLM_LDG(cos_ptr + x_index);
+    sin_f = VLLM_LDG(sin_ptr + x_index);
   } else {
-    // GPT-J style rotary embedding.
     x_index = 2 * rot_offset;
     y_index = 2 * rot_offset + 1;
-    cos = VLLM_LDG(cos_ptr + x_index / 2);
-    sin = VLLM_LDG(sin_ptr + x_index / 2);
+    cos_f = VLLM_LDG(cos_ptr + x_index / 2);
+    sin_f = VLLM_LDG(sin_ptr + x_index / 2);
   }
-
-  const scalar_t x = arr[x_index];
-  const scalar_t y = arr[y_index];
-  arr[x_index] = x * cos - y * sin;
-  arr[y_index] = y * cos + x * sin;
+  if (inverse) {
+    sin_f = -sin_f;
+  }
+  const float x_f = static_cast<float>(arr[x_index]);
+  const float y_f = static_cast<float>(arr[y_index]);
+  arr[x_index] = static_cast<scalar_t>(x_f * cos_f - y_f * sin_f);
+  arr[y_index] = static_cast<scalar_t>(y_f * cos_f + x_f * sin_f);
 }
 
 template <typename scalar_t, bool IS_NEOX>
@@ -42,22 +43,23 @@ inline __device__ void apply_rotary_embedding(
                                    // [batch_size, seq_len, num_kv_heads,
                                    // head_size] or [num_tokens, num_kv_heads,
                                    // head_size]
-    const scalar_t* cache_ptr, const int head_size, const int num_heads,
+    const float* cache_ptr, const int head_size, const int num_heads,
     const int num_kv_heads, const int rot_dim, const int token_idx,
     const int64_t query_stride, const int64_t key_stride,
-    const int64_t head_stride) {
+    const int64_t head_stride, const int64_t rope_dim_offset,
+    const bool inverse) {
   const int embed_dim = rot_dim / 2;
-  const scalar_t* cos_ptr = cache_ptr;
-  const scalar_t* sin_ptr = cache_ptr + embed_dim;
+  const float* cos_ptr = cache_ptr;
+  const float* sin_ptr = cache_ptr + embed_dim;
 
   const int nq = num_heads * embed_dim;
   for (int i = threadIdx.x; i < nq; i += blockDim.x) {
     const int head_idx = i / embed_dim;
     const int64_t token_head =
-        token_idx * query_stride + head_idx * head_stride;
+        token_idx * query_stride + head_idx * head_stride + rope_dim_offset;
     const int rot_offset = i % embed_dim;
     apply_token_rotary_embedding<scalar_t, IS_NEOX>(
-        query + token_head, cos_ptr, sin_ptr, rot_offset, embed_dim);
+        query + token_head, cos_ptr, sin_ptr, rot_offset, embed_dim, inverse);
   }
 
   if (key != nullptr) {
@@ -65,10 +67,10 @@ inline __device__ void apply_rotary_embedding(
     for (int i = threadIdx.x; i < nk; i += blockDim.x) {
       const int head_idx = i / embed_dim;
       const int64_t token_head =
-          token_idx * key_stride + head_idx * head_stride;
+          token_idx * key_stride + head_idx * head_stride + rope_dim_offset;
       const int rot_offset = i % embed_dim;
       apply_token_rotary_embedding<scalar_t, IS_NEOX>(
-          key + token_head, cos_ptr, sin_ptr, rot_offset, embed_dim);
+          key + token_head, cos_ptr, sin_ptr, rot_offset, embed_dim, inverse);
     }
   }
 }
@@ -84,19 +86,18 @@ __global__ void rotary_embedding_kernel(
                                  // [batch_size, seq_len, num_kv_heads,
                                  // head_size] or [num_tokens, num_kv_heads,
                                  // head_size]
-    const scalar_t* __restrict__ cos_sin_cache,  // [max_position, 2, rot_dim //
-                                                 // 2]
+    const float* __restrict__ cos_sin_cache,  // [max_position, rot_dim] fp32
     const int rot_dim, const int64_t query_stride, const int64_t key_stride,
     const int64_t head_stride, const int num_heads, const int num_kv_heads,
-    const int head_size) {
-  // Each thread block is responsible for one token.
+    const int head_size, const int64_t rope_dim_offset, const bool inverse) {
   const int token_idx = blockIdx.x;
   int64_t pos = positions[token_idx];
-  const scalar_t* cache_ptr = cos_sin_cache + pos * rot_dim;
+  const float* cache_ptr = cos_sin_cache + pos * rot_dim;
 
   apply_rotary_embedding<scalar_t, IS_NEOX>(
       query, key, cache_ptr, head_size, num_heads, num_kv_heads, rot_dim,
-      token_idx, query_stride, key_stride, head_stride);
+      token_idx, query_stride, key_stride, head_stride, rope_dim_offset,
+      inverse);
 }
 
 }  // namespace vllm
@@ -115,7 +116,7 @@ void rotary_embedding(
     // [num_tokens, num_heads, head_size]
     int64_t head_size,
     torch::Tensor& cos_sin_cache,  // [max_position, rot_dim]
-    bool is_neox) {
+    bool is_neox, int64_t rope_dim_offset, bool inverse) {
   // num_tokens = batch_size * seq_len
   int64_t num_tokens = positions.numel();
   int positions_ndim = positions.dim();
@@ -154,6 +155,8 @@ void rotary_embedding(
   int seq_dim_idx = positions_ndim - 1;
   int64_t query_stride = query.stride(seq_dim_idx);
   int64_t key_stride = key.has_value() ? key->stride(seq_dim_idx) : 0;
+
+  TORCH_CHECK((rot_dim + rope_dim_offset) <= head_size);
   // Determine head stride: for [*, heads, head_size] use stride of last dim;
   // for flat [*, heads*head_size], heads blocks are contiguous of size
   // head_size
@@ -165,20 +168,23 @@ void rotary_embedding(
   dim3 block(std::min<int64_t>(num_heads * rot_dim / 2, 512));
   const at::cuda::OptionalCUDAGuard device_guard(device_of(query));
   const cudaStream_t stream = at::cuda::getCurrentCUDAStream();
+  auto cache_f32 = cos_sin_cache.to(torch::kFloat32);
   VLLM_DISPATCH_FLOATING_TYPES(query.scalar_type(), "rotary_embedding", [&] {
     if (is_neox) {
       vllm::rotary_embedding_kernel<scalar_t, true><<<grid, block, 0, stream>>>(
           positions.data_ptr<int64_t>(), query.data_ptr<scalar_t>(),
           key.has_value() ? key->data_ptr<scalar_t>() : nullptr,
-          cos_sin_cache.data_ptr<scalar_t>(), rot_dim, query_stride, key_stride,
-          head_stride, num_heads, num_kv_heads, head_size);
+          cache_f32.data_ptr<float>(), rot_dim, query_stride, key_stride,
+          head_stride, num_heads, num_kv_heads, head_size, rope_dim_offset,
+          inverse);
     } else {
       vllm::rotary_embedding_kernel<scalar_t, false>
           <<<grid, block, 0, stream>>>(
               positions.data_ptr<int64_t>(), query.data_ptr<scalar_t>(),
               key.has_value() ? key->data_ptr<scalar_t>() : nullptr,
-              cos_sin_cache.data_ptr<scalar_t>(), rot_dim, query_stride,
-              key_stride, head_stride, num_heads, num_kv_heads, head_size);
+              cache_f32.data_ptr<float>(), rot_dim, query_stride, key_stride,
+              head_stride, num_heads, num_kv_heads, head_size, rope_dim_offset,
+              inverse);
     }
   });
 }

csrc/rocm/attention.cu

@@ -40,15 +40,6 @@ using __hip_fp8_e5m2 = __hip_fp8_e5m2_fnuz;
   #define __HIP__FP8MFMA__
 #endif
 
-#if defined(__HIPCC__) && (defined(__gfx1100__) || defined(__gfx1101__) || \
-                           defined(__gfx1150__) || defined(__gfx1151__))
-  #define __HIP__GFX11__
-#endif
-
-#if defined(__HIPCC__) && (defined(__gfx1200__) || defined(__gfx1201__))
-  #define __HIP__GFX12__
-#endif
-
 #if defined(NDEBUG)
   #undef NDEBUG
   #include <assert.h>
@@ -1629,7 +1620,7 @@ __launch_bounds__(NUM_THREADS) void paged_attention_ll4mi_reduce_kernel(
   }
 }
 
-#elif defined(__HIP__GFX11__)
+#elif defined(__GFX11__)
 
 using floatx8 = __attribute__((__vector_size__(8 * sizeof(float)))) float;
 
@@ -2388,7 +2379,7 @@ __launch_bounds__(NUM_THREADS) void paged_attention_ll4mi_reduce_kernel(
   out_ptr[threadIdx.x] = from_float<scalar_t>(acc);
 }
 
-#elif defined(__HIP__GFX12__)
+#elif defined(__GFX12__)
 
 using floatx8 = __attribute__((__vector_size__(8 * sizeof(float)))) float;
 

csrc/rocm/skinny_gemms.cu

@@ -26,16 +26,11 @@
   #define __HIP__GFX9__
 #endif
 
-#if defined(__HIPCC__) &&                                                    \
-    (defined(__gfx1100__) || defined(__gfx1101__) || defined(__gfx1150__) || \
-     defined(__gfx1151__) || defined(__gfx1200__) || defined(__gfx1201__))
+// Combined RDNA macro (gfx11 + gfx12) - both use 32-wide wavefronts
+#if defined(__GFX11__) || defined(__GFX12__)
   #define __HIP__GFX1X__
 #endif
 
-#if defined(__HIPCC__) && (defined(__gfx1200__) || defined(__gfx1201__))
-  #define __HIP__GFX12__
-#endif
-
 #if defined(__HIPCC__) && (defined(__gfx942__) || defined(__gfx950__))
   #define __HIP__MI3XX__
 #endif
@@ -1845,7 +1840,7 @@ torch::Tensor wvSplitKrc(const at::Tensor& in_a, const at::Tensor& in_b,
   return out_c;
 }
 
-#if defined(__HIP__MI3XX__) || defined(__HIP__GFX12__)
+#if defined(__HIP__MI3XX__) || defined(__GFX12__)
 template <typename scalar_t, typename fp8_t, int THRDS, int YTILE, int WvPrGrp,
           int A_CHUNK, int UNRL, int N>
 __global__ void __launch_bounds__(WvPrGrp* THRDS)
@@ -1893,7 +1888,7 @@ __global__ void __launch_bounds__(WvPrGrp* THRDS)
   float sB = *s_B;
 
   while (m < M) {
-  #ifdef __HIP__GFX12__
+  #ifdef __GFX12__
     // gfx12: per-lane scalar accumulation via v_dot4_f32_fp8_fp8
     float sum[N][YTILE] = {};
   #else
@@ -1931,7 +1926,7 @@ __global__ void __launch_bounds__(WvPrGrp* THRDS)
   #pragma unroll
       for (uint32_t k2 = 0; k2 < UNRL; k2++) {
         for (uint32_t n = 0; n < N; n++) {
-  #ifdef __HIP__GFX12__
+  #ifdef __GFX12__
           // gfx12: 4 x dot4 per A_CHUNK=16 bytes (4 FP8 per dot4)
           for (int y = 0; y < YTILE; ++y) {
     #pragma unroll
@@ -1955,7 +1950,7 @@ __global__ void __launch_bounds__(WvPrGrp* THRDS)
     }
 
     // Final reduction
-  #ifdef __HIP__GFX12__
+  #ifdef __GFX12__
     // gfx12 wave32: DPP row_shr within 16-lane rows + cross-row shuffle
     for (int n = 0; n < N; n++) {
       for (int y = 0; y < YTILE; y++) {
@@ -1993,7 +1988,7 @@ __global__ void __launch_bounds__(WvPrGrp* THRDS)
   #endif
 
     const bool writeback_lane =
-  #ifdef __HIP__GFX12__
+  #ifdef __GFX12__
         threadIdx.x == (THRDS - 1);
   #else
         threadIdx.x == 0;
@@ -2009,7 +2004,7 @@ __global__ void __launch_bounds__(WvPrGrp* THRDS)
       for (int n = 0; n < N; n++) {
         for (int y = 0; y < YTILE; y++) {
           if (y + m >= M) break;  // To avoid mem access fault.
-  #ifdef __HIP__GFX12__
+  #ifdef __GFX12__
           float result = sum[n][y] * sA * sB;
   #else
           float result = sum[n][y][0] * sA * sB;
@@ -2027,7 +2022,7 @@ __global__ void __launch_bounds__(WvPrGrp* THRDS)
     m += CuCount * _WvPrGrp * YTILE;
   }
 }
-#else   // !defined(__HIP__MI3XX__) && !defined(__HIP__GFX12__)
+#else   // !defined(__HIP__MI3XX__) && !defined(__GFX12__)
 template <typename scalar_t, typename fp8_t, int THRDS, int YTILE, int WvPrGrp,
           int A_CHUNK, int UNRL, int N>
 __global__ void wvSplitKQ_hf_sml_(const int K, const int Kap, const int Kbp,
@@ -2039,9 +2034,9 @@ __global__ void wvSplitKQ_hf_sml_(const int K, const int Kap, const int Kbp,
                                   const int _WvPrGrp, const int CuCount) {
   UNREACHABLE_CODE
 }
-#endif  // defined(__HIP__MI3XX__) || defined(__HIP__GFX12__)
+#endif  // defined(__HIP__MI3XX__) || defined(__GFX12__)
 
-#if defined(__HIP__MI3XX__) || defined(__HIP__GFX12__)
+#if defined(__HIP__MI3XX__) || defined(__GFX12__)
 template <typename scalar_t, typename fp8_t, int THRDS, int YTILE, int WvPrGrp,
           int A_CHUNK, int UNRL, int N>
 __global__ void __launch_bounds__(WvPrGrp* THRDS)
@@ -2088,7 +2083,7 @@ __global__ void __launch_bounds__(WvPrGrp* THRDS)
   float sB = *s_B;
 
   while (m < M) {
-  #ifdef __HIP__GFX12__
+  #ifdef __GFX12__
     // gfx12: per-lane scalar accumulation via v_dot4_f32_fp8_fp8
     float sum[N][YTILE] = {};
   #else
@@ -2128,7 +2123,7 @@ __global__ void __launch_bounds__(WvPrGrp* THRDS)
   #pragma unroll
       for (uint32_t k2 = 0; k2 < UNRL; k2++) {
         for (uint32_t n = 0; n < N; n++) {
-  #ifdef __HIP__GFX12__
+  #ifdef __GFX12__
           // gfx12: 4 x dot4 per A_CHUNK=16 bytes (4 FP8 per dot4)
           for (int y = 0; y < YTILE; ++y) {
     #pragma unroll
@@ -2152,7 +2147,7 @@ __global__ void __launch_bounds__(WvPrGrp* THRDS)
     }
 
     // Final reduction
-  #ifdef __HIP__GFX12__
+  #ifdef __GFX12__
     // gfx12 wave32: DPP row_shr within 16-lane rows + cross-row shuffle
     for (int n = 0; n < N; n++) {
       for (int y = 0; y < YTILE; y++) {
@@ -2190,7 +2185,7 @@ __global__ void __launch_bounds__(WvPrGrp* THRDS)
   #endif
 
     const bool writeback_lane =
-  #ifdef __HIP__GFX12__
+  #ifdef __GFX12__
         threadIdx.x == (THRDS - 1);
   #else
         threadIdx.x == 0;
@@ -2206,7 +2201,7 @@ __global__ void __launch_bounds__(WvPrGrp* THRDS)
       for (int n = 0; n < N; n++) {
         for (int y = 0; y < YTILE; y++) {
           if (y + m >= M) break;  // To avoid mem access fault.
-  #ifdef __HIP__GFX12__
+  #ifdef __GFX12__
           float result = sum[n][y] * sA * sB;
   #else
           float result = sum[n][y][0] * sA * sB;
@@ -2224,7 +2219,7 @@ __global__ void __launch_bounds__(WvPrGrp* THRDS)
     m += CuCount * _WvPrGrp * YTILE;
   }
 }
-#else   // !defined(__HIP__MI3XX__) && !defined(__HIP__GFX12__)
+#else   // !defined(__HIP__MI3XX__) && !defined(__GFX12__)
 template <typename scalar_t, typename fp8_t, int THRDS, int YTILE, int WvPrGrp,
           int A_CHUNK, int UNRL, int N>
 __global__ void wvSplitKQ_hf_(const int K, const int Kap, const int Kbp,
@@ -2236,7 +2231,7 @@ __global__ void wvSplitKQ_hf_(const int K, const int Kap, const int Kbp,
                               const int CuCount) {
   UNREACHABLE_CODE
 }
-#endif  // defined(__HIP__MI3XX__) || defined(__HIP__GFX12__)
+#endif  // defined(__HIP__MI3XX__) || defined(__GFX12__)
 
 void wvSplitKQ(const at::Tensor& in_b, const at::Tensor& in_a,
                const std::optional<at::Tensor>& in_bias, at::Tensor& out_c,

csrc/sampler.cu

@@ -258,7 +258,13 @@ __device__ bool processHistogramStep(
   auto processBins = [&](float logit, int idx) {
     if (isPartialMatch<patternShift>(logit, logitPattern)) {
       uint32_t binIdx = extractBinIdx<step>(logit);
-      if (binIdx < thresholdBinIdx) {
+      // Only write elements with binIdx < thresholdBinIdx when:
+      // 1. This is step 0 and the threshold bin is small enough (no step 1)
+      // 2. This is step >= 1 (where pattern matching filters correctly)
+      // This prevents duplicates when step 0 and step 1 both run.
+      bool shouldWriteDirectly =
+          (step == 0 && smemFinalBinSize[0] <= kNumFinalItems) || (step >= 1);
+      if (binIdx < thresholdBinIdx && shouldWriteDirectly) {
         // The element is part of the top-k selection
         int dstIdx = atomicAdd(&smemFoundTopKValues[0], 1);
 

csrc/topk.cu

@@ -10,33 +10,17 @@
   #include "persistent_topk.cuh"
 #endif
 
-void persistent_topk(const torch::Tensor& logits, const torch::Tensor& lengths,
-                     torch::Tensor& output, torch::Tensor& workspace, int64_t k,
-                     int64_t max_seq_len) {
+namespace {
+
 #ifndef USE_ROCM
-  TORCH_CHECK(logits.is_cuda(), "logits must be CUDA tensor");
-  TORCH_CHECK(lengths.is_cuda(), "lengths must be CUDA tensor");
-  TORCH_CHECK(output.is_cuda(), "output must be CUDA tensor");
-  TORCH_CHECK(logits.dtype() == torch::kFloat32, "Only float32 supported");
-  TORCH_CHECK(lengths.dtype() == torch::kInt32, "lengths must be int32");
-  TORCH_CHECK(output.dtype() == torch::kInt32, "output must be int32");
-  TORCH_CHECK(logits.dim() == 2, "logits must be 2D");
-  TORCH_CHECK(lengths.dim() == 1 || lengths.dim() == 2,
-              "lengths must be 1D or 2D");
-  TORCH_CHECK(lengths.is_contiguous(), "lengths must be contiguous");
-  TORCH_CHECK(output.dim() == 2, "output must be 2D");
+template <int TopK>
+void launch_persistent_topk(const torch::Tensor& logits,
+                            const torch::Tensor& lengths, torch::Tensor& output,
+                            torch::Tensor& workspace, int64_t max_seq_len) {
+  namespace P = vllm::persistent;
 
   const int64_t num_rows = logits.size(0);
   const int64_t stride = logits.size(1);
-
-  TORCH_CHECK(lengths.numel() == num_rows, "lengths size mismatch");
-  TORCH_CHECK(output.size(0) == num_rows && output.size(1) == k,
-              "output size mismatch");
-  namespace P = vllm::persistent;
-
-  TORCH_CHECK(k == P::TopK, "k must be 2048");
-  TORCH_CHECK(k <= stride, "k out of range");
-
   cudaStream_t stream = at::cuda::getCurrentCUDAStream();
 
   static int num_sms = 0;
@@ -50,18 +34,17 @@ void persistent_topk(const torch::Tensor& logits, const torch::Tensor& lengths,
   }
 
   if (num_rows > 32 && max_smem_per_block >= 128 * 1024) {
-    cudaError_t status = vllm::FilteredTopKRaggedTransform<float, int32_t>(
-        logits.data_ptr<float>(), output.data_ptr<int32_t>(),
-        lengths.data_ptr<int32_t>(), static_cast<uint32_t>(num_rows),
-        static_cast<uint32_t>(k), static_cast<uint32_t>(stride), stream);
+    cudaError_t status =
+        vllm::FilteredTopKRaggedTransform<float, int32_t, TopK>(
+            logits.data_ptr<float>(), output.data_ptr<int32_t>(),
+            lengths.data_ptr<int32_t>(), static_cast<uint32_t>(num_rows),
+            static_cast<uint32_t>(TopK), static_cast<uint32_t>(stride), stream);
     TORCH_CHECK(status == cudaSuccess,
                 "FilteredTopK failed: ", cudaGetErrorString(status));
   } else {
     TORCH_CHECK(workspace.is_cuda(), "workspace must be CUDA tensor");
     TORCH_CHECK(workspace.dtype() == torch::kUInt8, "workspace must be uint8");
 
-    // Smem cap: smaller smem → more CTAs/group → more per-row parallelism for
-    // large path. Empirically tuned.
     int effective_max_smem;
     if (num_rows <= 4) {
       effective_max_smem =
@@ -101,7 +84,7 @@ void persistent_topk(const torch::Tensor& logits, const torch::Tensor& lengths,
 
     int occupancy = 1;
     cudaOccupancyMaxActiveBlocksPerMultiprocessor(
-        &occupancy, P::persistent_topk_kernel<4>, P::kThreadsPerBlock,
+        &occupancy, P::persistent_topk_kernel<TopK, 4>, P::kThreadsPerBlock,
         smem_size);
     if (occupancy < 1) occupancy = 1;
 
@@ -121,15 +104,16 @@ void persistent_topk(const torch::Tensor& logits, const torch::Tensor& lengths,
     params.lengths = lengths.data_ptr<int32_t>();
     params.num_rows = static_cast<uint32_t>(num_rows);
     params.stride = static_cast<uint32_t>(stride);
+    params.top_k = static_cast<uint32_t>(TopK);
     params.chunk_size = chunk_size;
     params.row_states =
         reinterpret_cast<P::RadixRowState*>(workspace.data_ptr<uint8_t>());
     params.ctas_per_group = ctas_per_group;
     params.max_seq_len = static_cast<uint32_t>(max_seq_len);
 
-  #define LAUNCH_PERSISTENT(VS)                                               \
+  #define LAUNCH_PERSISTENT(TOPK_VAL, VS)                                     \
     do {                                                                      \
-      auto kernel = &P::persistent_topk_kernel<VS>;                           \
+      auto kernel = &P::persistent_topk_kernel<TOPK_VAL, VS>;                 \
       cudaError_t err = cudaFuncSetAttribute(                                 \
           kernel, cudaFuncAttributeMaxDynamicSharedMemorySize, smem_size);    \
       TORCH_CHECK(err == cudaSuccess,                                         \
@@ -138,11 +122,11 @@ void persistent_topk(const torch::Tensor& logits, const torch::Tensor& lengths,
     } while (0)
 
     if (vec_size == 4) {
-      LAUNCH_PERSISTENT(4);
+      LAUNCH_PERSISTENT(TopK, 4);
     } else if (vec_size == 2) {
-      LAUNCH_PERSISTENT(2);
+      LAUNCH_PERSISTENT(TopK, 2);
     } else {
-      LAUNCH_PERSISTENT(1);
+      LAUNCH_PERSISTENT(TopK, 1);
     }
   #undef LAUNCH_PERSISTENT
   }
@@ -150,6 +134,46 @@ void persistent_topk(const torch::Tensor& logits, const torch::Tensor& lengths,
   cudaError_t err = cudaGetLastError();
   TORCH_CHECK(err == cudaSuccess,
               "persistent_topk failed: ", cudaGetErrorString(err));
+}
+#endif
+
+}  // anonymous namespace
+
+void persistent_topk(const torch::Tensor& logits, const torch::Tensor& lengths,
+                     torch::Tensor& output, torch::Tensor& workspace, int64_t k,
+                     int64_t max_seq_len) {
+#ifndef USE_ROCM
+  TORCH_CHECK(logits.is_cuda(), "logits must be CUDA tensor");
+  TORCH_CHECK(lengths.is_cuda(), "lengths must be CUDA tensor");
+  TORCH_CHECK(output.is_cuda(), "output must be CUDA tensor");
+  TORCH_CHECK(logits.dtype() == torch::kFloat32, "Only float32 supported");
+  TORCH_CHECK(lengths.dtype() == torch::kInt32, "lengths must be int32");
+  TORCH_CHECK(output.dtype() == torch::kInt32, "output must be int32");
+  TORCH_CHECK(logits.dim() == 2, "logits must be 2D");
+  TORCH_CHECK(lengths.dim() == 1 || lengths.dim() == 2,
+              "lengths must be 1D or 2D");
+  TORCH_CHECK(lengths.is_contiguous(), "lengths must be contiguous");
+  TORCH_CHECK(output.dim() == 2, "output must be 2D");
+
+  const int64_t num_rows = logits.size(0);
+  const int64_t stride = logits.size(1);
+
+  TORCH_CHECK(lengths.numel() == num_rows, "lengths size mismatch");
+  TORCH_CHECK(output.size(0) == num_rows && output.size(1) == k,
+              "output size mismatch");
+  TORCH_CHECK(k == 512 || k == 1024 || k == 2048,
+              "persistent_topk supports k=512, k=1024, or k=2048, got k=", k);
+
+  if (k == 512) {
+    launch_persistent_topk<512>(logits, lengths, output, workspace,
+                                max_seq_len);
+  } else if (k == 1024) {
+    launch_persistent_topk<1024>(logits, lengths, output, workspace,
+                                 max_seq_len);
+  } else {
+    launch_persistent_topk<2048>(logits, lengths, output, workspace,
+                                 max_seq_len);
+  }
 #else
   TORCH_CHECK(false, "persistent_topk is not supported on ROCm");
 #endif

csrc/torch_bindings.cpp

@@ -106,6 +106,12 @@ TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, ops) {
   ops.def("silu_and_mul(Tensor! result, Tensor input) -> ()");
   ops.impl("silu_and_mul", torch::kCUDA, &silu_and_mul);
 
+  // SwiGLU activation with input clamping.
+  ops.def(
+      "silu_and_mul_with_clamp(Tensor! result, Tensor input, float limit) "
+      "-> ()");
+  ops.impl("silu_and_mul_with_clamp", torch::kCUDA, &silu_and_mul_clamp);
+
   ops.def(
       "silu_and_mul_quant(Tensor! result, Tensor input, Tensor scale) -> ()");
   ops.impl("silu_and_mul_quant", torch::kCUDA, &silu_and_mul_quant);
@@ -177,6 +183,19 @@ TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, ops) {
       "int forced_token_heads_per_warp=-1) -> ()");
   ops.impl("fused_qk_norm_rope", torch::kCUDA, &fused_qk_norm_rope);
 
+#ifndef USE_ROCM
+  // Horizontally-fused DeepseekV4-MLA: per-head RMSNorm + GPT-J RoPE for Q, and
+  // GPT-J RoPE + UE8M0 FP8 quant + paged cache insert for KV, all in one
+  // kernel launch.
+  ops.def(
+      "fused_deepseek_v4_qnorm_rope_kv_rope_quant_insert("
+      "Tensor! q, Tensor kv, Tensor! k_cache, "
+      "Tensor slot_mapping, Tensor position_ids, Tensor cos_sin_cache, "
+      "float eps, int cache_block_size) -> ()");
+  ops.impl("fused_deepseek_v4_qnorm_rope_kv_rope_quant_insert", torch::kCUDA,
+           &fused_deepseek_v4_qnorm_rope_kv_rope_quant_insert);
+#endif
+
   // Apply repetition penalties to logits in-place
   ops.def(
       "apply_repetition_penalties_(Tensor! logits, Tensor prompt_mask, "
@@ -240,7 +259,8 @@ TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, ops) {
   ops.def(
       "rotary_embedding(Tensor positions, Tensor! query,"
       "                 Tensor!? key, int head_size,"
-      "                 Tensor cos_sin_cache, bool is_neox) -> ()");
+      "                 Tensor cos_sin_cache, bool is_neox, int "
+      "rope_dim_offset=0, bool inverse=False) -> ()");
   ops.impl("rotary_embedding", torch::kCUDA, &rotary_embedding);
 
   // Quantization ops

docker/Dockerfile

@@ -22,7 +22,7 @@
 #   docker buildx bake -f docker/docker-bake.hcl -f docker/versions.json
 # =============================================================================
 
-ARG CUDA_VERSION=13.0.0
+ARG CUDA_VERSION=13.0.2
 ARG PYTHON_VERSION=3.12
 ARG UBUNTU_VERSION=22.04
 
@@ -188,7 +188,7 @@ RUN --mount=type=cache,target=/root/.cache/uv \
 # Explicitly set the list to avoid issues with torch 2.2
 # See https://github.com/pytorch/pytorch/pull/123243
 # From versions.json: .torch.cuda_arch_list
-ARG torch_cuda_arch_list='7.0 7.5 8.0 8.9 9.0 10.0 12.0'
+ARG torch_cuda_arch_list='7.5 8.0 8.6 8.9 9.0 10.0 11.0 12.0+PTX'
 ENV TORCH_CUDA_ARCH_LIST=${torch_cuda_arch_list}
 #################### BUILD BASE IMAGE ####################
 
@@ -538,9 +538,11 @@ RUN CUDA_VERSION_DASH=$(echo $CUDA_VERSION | cut -d. -f1,2 | tr '.' '-') && \
         cuda-nvrtc-${CUDA_VERSION_DASH} \
         cuda-cuobjdump-${CUDA_VERSION_DASH} \
         libcurand-dev-${CUDA_VERSION_DASH} \
-        libcublas-${CUDA_VERSION_DASH} \
+        libcublas-dev-${CUDA_VERSION_DASH} \
         # Required by fastsafetensors (fixes #20384)
-        libnuma-dev && \
+        libnuma-dev \
+        # numactl CLI for NUMA binding at runtime
+        numactl && \
     # Fixes nccl_allocator requiring nccl.h at runtime
     # https://github.com/vllm-project/vllm/blob/1336a1ea244fa8bfd7e72751cabbdb5b68a0c11a/vllm/distributed/device_communicators/pynccl_allocator.py#L22
     # NCCL packages don't use the cuda-MAJOR-MINOR naming convention,
@@ -580,33 +582,10 @@ RUN --mount=type=cache,target=/root/.cache/uv \
 # Install FlashInfer JIT cache (requires CUDA-version-specific index URL)
 # https://docs.flashinfer.ai/installation.html
 # From versions.json: .flashinfer.version
-# 0.6.7: CUTLASS 4.4.2 bump, fixes TMA grouped GEMM on SM12x (flashinfer#2798)
-# TODO: bump to 0.6.8 when released for NVFP4/MXFP4 group GEMMs on
-#   SM120/SM121 (RTX 50 / DGX Spark) via flashinfer#2738
-ARG FLASHINFER_VERSION=0.6.7
+ARG FLASHINFER_VERSION=0.6.8.post1
 RUN --mount=type=cache,target=/root/.cache/uv \
     uv pip install --system flashinfer-jit-cache==${FLASHINFER_VERSION} \
-        --extra-index-url https://flashinfer.ai/whl/cu$(echo $CUDA_VERSION | cut -d. -f1,2 | tr -d '.') \
-    && flashinfer show-config
-
-# Pre-download FlashInfer TRTLLM BMM headers for air-gapped environments.
-# At runtime, MoE JIT compilation downloads these from edge.urm.nvidia.com
-# which fails without internet. This step caches them at build time.
-RUN python3 <<'PYEOF'
-from flashinfer.jit import env as jit_env
-from flashinfer.jit.cubin_loader import download_trtllm_headers, get_cubin
-from flashinfer.artifacts import ArtifactPath, CheckSumHash
-
-download_trtllm_headers(
-    'bmm',
-    jit_env.FLASHINFER_CUBIN_DIR / 'flashinfer' / 'trtllm' / 'batched_gemm' / 'trtllmGen_bmm_export',
-    f'{ArtifactPath.TRTLLM_GEN_BMM}/include/trtllmGen_bmm_export',
-    ArtifactPath.TRTLLM_GEN_BMM,
-    get_cubin(f'{ArtifactPath.TRTLLM_GEN_BMM}/checksums.txt', CheckSumHash.TRTLLM_GEN_BMM),
-)
-
-print('FlashInfer TRTLLM BMM headers downloaded successfully')
-PYEOF
+        --extra-index-url https://flashinfer.ai/whl/cu$(echo $CUDA_VERSION | cut -d. -f1,2 | tr -d '.')
 
 # ============================================================
 # OPENAI API SERVER DEPENDENCIES
@@ -688,6 +667,13 @@ RUN --mount=type=bind,from=build,src=/tmp/ep_kernels_workspace/dist,target=/vllm
     uv pip install --system ep_kernels/dist/*.whl --verbose \
         --extra-index-url ${PYTORCH_CUDA_INDEX_BASE_URL}/cu$(echo $CUDA_VERSION | cut -d. -f1,2 | tr -d '.')
 
+# Download FlashInfer precompiled cubins AFTER all pip installs are done.
+# This must run after the vLLM wheel and EP kernels installs above, because
+# those can reinstall/touch flashinfer packages. Downloading cubins earlier
+# (in the flashinfer-jit-cache layer) causes ~2.5 GB of layer duplication
+# when a later pip install overwrites flashinfer package files.
+RUN flashinfer show-config && flashinfer download-cubin
+
 # CUDA image changed from /usr/local/nvidia to /usr/local/cuda in 12.8 but will
 # return to /usr/local/nvidia in 13.0 to allow container providers to mount drivers
 # consistently from the host (see https://github.com/vllm-project/vllm/issues/18859).
@@ -786,7 +772,7 @@ ARG PIP_EXTRA_INDEX_URL UV_EXTRA_INDEX_URL
 ENV UV_HTTP_TIMEOUT=500
 
 # install kv_connectors if requested
-ARG torch_cuda_arch_list='7.0 7.5 8.0 8.9 9.0 10.0 12.0'
+ARG torch_cuda_arch_list='7.5 8.0 8.6 8.9 9.0 10.0 11.0 12.0+PTX'
 ENV TORCH_CUDA_ARCH_LIST=${torch_cuda_arch_list}
 RUN --mount=type=cache,target=/root/.cache/uv \
     --mount=type=bind,source=requirements/kv_connectors.txt,target=/tmp/kv_connectors.txt,ro \

docker/Dockerfile.nightly_torch

@@ -77,7 +77,7 @@ RUN --mount=type=cache,target=/root/.cache/uv \
     uv pip install --system $pkgs --index-url https://download.pytorch.org/whl/nightly/cu128
 
 RUN --mount=type=cache,target=/root/.cache/uv \
-    uv pip install --system numba==0.61.2
+    uv pip install --system numba==0.65.0
 
 RUN --mount=type=cache,target=/root/.cache/uv \
 uv pip install --system -r requirements/common.txt
@@ -217,16 +217,13 @@ RUN pip install setuptools==75.6.0 packaging==23.2 ninja==1.11.1.3 build==1.2.2.
 
 
 # build flashinfer for torch nightly from source around 10 mins
-# release version: v0.6.7
-# 0.6.7: CUTLASS 4.4.2 bump, fixes TMA grouped GEMM on SM12x (flashinfer#2798)
-# TODO: bump to 0.6.8 when released for NVFP4/MXFP4 group GEMMs on
-#   SM120/SM121 (RTX 50 / DGX Spark) via flashinfer#2738
+# release version: v0.6.8.post1
 # todo(elainewy): cache flashinfer build result for faster build
 ENV CCACHE_DIR=/root/.cache/ccache
 RUN --mount=type=cache,target=/root/.cache/ccache \
     --mount=type=cache,target=/root/.cache/uv \
     echo "git clone flashinfer..." \
-    && git clone --depth 1 --branch v0.6.7 --recursive https://github.com/flashinfer-ai/flashinfer.git \
+    && git clone --depth 1 --branch v0.6.8.post1 --recursive https://github.com/flashinfer-ai/flashinfer.git \
     && cd flashinfer \
     && git submodule update --init --recursive \
     && echo "finish git clone flashinfer..." \

docker/Dockerfile.rocm

@@ -2,6 +2,11 @@
 ARG REMOTE_VLLM="0"
 ARG COMMON_WORKDIR=/app
 ARG BASE_IMAGE=rocm/vllm-dev:base
+# AMD NIC backend
+ARG NIC_BACKEND=none
+# AMD AINIC apt repo settings
+ARG AINIC_VERSION=1.117.5
+ARG UBUNTU_CODENAME=jammy
 
 # Sccache configuration (only used in release pipeline)
 ARG USE_SCCACHE
@@ -119,10 +124,10 @@ COPY --from=build_vllm ${COMMON_WORKDIR}/vllm/vllm/v1 /vllm_v1
 
 # RIXL/UCX build stages
 FROM base AS build_rixl
-ARG RIXL_BRANCH="f33a5599"
+ARG RIXL_BRANCH="bf4a7214"
 ARG RIXL_REPO="https://github.com/ROCm/RIXL.git"
-ARG UCX_BRANCH="da3fac2a"
-ARG UCX_REPO="https://github.com/ROCm/ucx.git"
+ARG UCX_BRANCH="7009d7a1"
+ARG UCX_REPO="https://github.com/openucx/ucx.git"
 ENV ROCM_PATH=/opt/rocm
 ENV UCX_HOME=/usr/local/ucx
 ENV RIXL_HOME=/usr/local/rixl
@@ -160,7 +165,7 @@ RUN cd /usr/local/src && \
         --disable-doxygen-doc \
         --enable-optimizations \
         --enable-devel-headers \
-        --with-rocm=/opt/rocm \
+        --with-rocm=${ROCM_PATH} \
         --with-verbs \
         --with-dm \
         --enable-mt && \
@@ -181,7 +186,12 @@ RUN git clone ${RIXL_REPO} /opt/rixl && \
     ninja install
 
 # Generate RIXL wheel
-RUN cd /opt/rixl && mkdir -p /app/install && \
+# Exclude libcore and libpull from auditwheel: transitive dependencies
+# that are not shipped in the wheel and vary across base images.
+RUN cd /opt/rixl && \
+    sed -i "s/--exclude 'libamdhip64\*'/--exclude 'libamdhip64*' --exclude 'libcore*' --exclude 'libpull*'/" \
+        contrib/build-wheel.sh && \
+    mkdir -p /app/install && \
     ./contrib/build-wheel.sh \
         --output-dir /app/install \
         --rocm-dir ${ROCM_PATH} \
@@ -192,9 +202,10 @@ RUN cd /opt/rixl && mkdir -p /app/install && \
 FROM base AS build_deep
 ARG ROCSHMEM_BRANCH="ba0bf0f3"
 ARG ROCSHMEM_REPO="https://github.com/ROCm/rocm-systems.git"
-ARG DEEPEP_BRANCH="e84464ec"
+ARG DEEPEP_BRANCH="5d90af8b"
 ARG DEEPEP_REPO="https://github.com/ROCm/DeepEP.git"
 ARG DEEPEP_NIC="cx7"
+ARG DEEPEP_ROCM_ARCH="gfx942;gfx950"
 ENV ROCSHMEM_DIR=/opt/rocshmem
 
 RUN git clone ${ROCSHMEM_REPO} \
@@ -202,13 +213,11 @@ RUN git clone ${ROCSHMEM_REPO} \
  && git checkout ${ROCSHMEM_BRANCH} \
  && mkdir -p projects/rocshmem/build \
  && cd projects/rocshmem/build \
- && cmake .. \
-    -DCMAKE_INSTALL_PREFIX="${ROCSHMEM_DIR}" \
-    -DROCM_PATH=/opt/rocm \
-    -DCMAKE_POSITION_INDEPENDENT_CODE=ON \
-    -DUSE_EXTERNAL_MPI=OFF \
- && make -j \
- && make install
+ && bash ../scripts/build_configs/all_backends \
+      -DCMAKE_INSTALL_PREFIX="${ROCSHMEM_DIR}" \
+      -DROCM_PATH=/opt/rocm \
+      -DGPU_TARGETS="${DEEPEP_ROCM_ARCH}" \
+      -DUSE_EXTERNAL_MPI=OFF
 
 # Build DeepEP wheel.
 # DeepEP looks for rocshmem at ROCSHMEM_DIR.
@@ -217,6 +226,47 @@ RUN git clone ${DEEPEP_REPO} \
  && git checkout ${DEEPEP_BRANCH} \
  && python3 setup.py --variant rocm --nic ${DEEPEP_NIC} bdist_wheel --dist-dir=/app/deep_install
 
+# MoRI runtime dependencies live in Dockerfile.rocm so NIC backend changes do
+# not force users to rebuild the long-lived Dockerfile.rocm_base image.
+FROM base AS mori_base
+ARG NIC_BACKEND
+ARG AINIC_VERSION
+ARG UBUNTU_CODENAME
+RUN /bin/bash -lc 'set -euo pipefail; \
+ echo "[MORI] Install MoRI proxy deps"; \
+ pip install --quiet --ignore-installed blinker && \
+ pip install --quiet quart msgpack aiohttp pyzmq; \
+ echo "[MORI] NIC_BACKEND=${NIC_BACKEND}"; \
+  \
+  # NIC backend deps — mori auto-detects NIC at runtime (MORI_DEVICE_NIC env var override).
+  # Only vendor packages are installed here for dlopen (e.g. libionic.so); no compile-time flags needed.
+  case "${NIC_BACKEND}" in \
+    # default: mlx5
+    none) \
+      ;; \
+    # AMD NIC
+    ainic) \
+      apt-get update && apt-get install -y --no-install-recommends ca-certificates curl gnupg apt-transport-https && \
+      rm -rf /var/lib/apt/lists/* && mkdir -p /etc/apt/keyrings; \
+      curl -fsSL https://repo.radeon.com/rocm/rocm.gpg.key | gpg --dearmor > /etc/apt/keyrings/amdainic.gpg; \
+      echo "deb [arch=amd64 signed-by=/etc/apt/keyrings/amdainic.gpg] https://repo.radeon.com/amdainic/pensando/ubuntu/${AINIC_VERSION} ${UBUNTU_CODENAME} main" \
+        > /etc/apt/sources.list.d/amdainic.list; \
+      apt-get update && apt-get install -y --no-install-recommends \
+          libionic-dev \
+          ionic-common \
+      ; \
+      rm -rf /var/lib/apt/lists/*; \
+      ;; \
+    # TODO: Add Broadcom bnxt packages/repos here later.
+    # bnxt) \
+    #   echo "[MORI] Add Broadcom bnxt packages/repos here later."; \
+    #   ;; \
+    *) \
+      echo "ERROR: unknown NIC_BACKEND=${NIC_BACKEND}. Use one of: none, ainic"; \
+      exit 2; \
+      ;; \
+  esac;'
+
 # -----------------------
 # vLLM wheel release build stage (for building distributable wheels)
 # This stage pins dependencies to custom ROCm wheel versions and handles version detection
@@ -319,7 +369,7 @@ COPY --from=build_vllm_wheel_release ${COMMON_WORKDIR}/vllm/vllm/v1 /vllm_v1
 
 # -----------------------
 # Test vLLM image
-FROM base AS test
+FROM mori_base AS test
 
 RUN python3 -m pip install --upgrade pip && rm -rf /var/lib/apt/lists/*
 
@@ -386,6 +436,10 @@ COPY --from=export_vllm /vllm_v1 /usr/local/lib/python${PYTHON_VERSION}/dist-pac
 ENV MIOPEN_DEBUG_CONV_DIRECT=0
 ENV MIOPEN_DEBUG_CONV_GEMM=0
 
+# Use legacy IPC mode for HSA to avoid GPU memory pinning issues with UCX rocm_ipc
+# See: https://github.com/ROCm/rocm-libraries/issues/6266
+ENV HSA_ENABLE_IPC_MODE_LEGACY=1
+
 # Source code is used in the `python_only_compile.sh` test
 # We hide it inside `src/` so that this source code
 # will not be imported by other tests
@@ -410,7 +464,7 @@ RUN printf '%s\n' \
 
 # -----------------------
 # Final vLLM image
-FROM base AS final
+FROM mori_base AS final
 
 RUN python3 -m pip install --upgrade pip && rm -rf /var/lib/apt/lists/*
 
@@ -447,6 +501,8 @@ RUN --mount=type=bind,from=export_vllm,src=/,target=/install \
 
 ARG COMMON_WORKDIR
 ARG BASE_IMAGE
+ARG NIC_BACKEND
+ARG AINIC_VERSION
 
 # Copy over the benchmark scripts as well
 COPY --from=export_vllm /benchmarks ${COMMON_WORKDIR}/vllm/benchmarks
@@ -464,7 +520,9 @@ ENV HIP_FORCE_DEV_KERNARG=1
 # Workaround for ROCm profiler limits
 RUN echo "ROCTRACER_MAX_EVENTS=10000000" > ${COMMON_WORKDIR}/libkineto.conf
 ENV KINETO_CONFIG="${COMMON_WORKDIR}/libkineto.conf"
-RUN echo "VLLM_BASE_IMAGE=${BASE_IMAGE}" >> ${COMMON_WORKDIR}/versions.txt
+RUN echo "VLLM_BASE_IMAGE=${BASE_IMAGE}" >> ${COMMON_WORKDIR}/versions.txt \
+    && echo "MORI_NIC_BACKEND=${NIC_BACKEND}" >> ${COMMON_WORKDIR}/versions.txt \
+    && echo "AINIC_VERSION=${AINIC_VERSION}" >> ${COMMON_WORKDIR}/versions.txt
 
 CMD ["/bin/bash"]
 

docker/Dockerfile.rocm_base

@@ -11,7 +11,7 @@ ARG FA_BRANCH="0e60e394"
 ARG FA_REPO="https://github.com/Dao-AILab/flash-attention.git"
 ARG AITER_BRANCH="v0.1.10.post3"
 ARG AITER_REPO="https://github.com/ROCm/aiter.git"
-ARG MORI_BRANCH="2d02c6a9"
+ARG MORI_BRANCH="v1.1.0"
 ARG MORI_REPO="https://github.com/ROCm/mori.git"
 
 # Sccache configuration (only used in release pipeline)

docker/Dockerfile.xpu

@@ -50,9 +50,9 @@ RUN curl -LsSf https://astral.sh/uv/install.sh | sh
 RUN uv venv --python ${PYTHON_VERSION} --seed ${VIRTUAL_ENV}
 ENV PATH="$VIRTUAL_ENV/bin:$PATH"
 
-# This oneccl contains the BMG support which is not the case for default version of oneapi 2025.2.
-ARG ONECCL_INSTALLER="intel-oneccl-2021.15.7.8_offline.sh"
-RUN wget "https://github.com/uxlfoundation/oneCCL/releases/download/2021.15.7/${ONECCL_INSTALLER}" && \
+# This oneccl contains the BMG support which is not the case for default version of oneapi 2025.3.
+ARG ONECCL_INSTALLER="intel-oneccl-2021.15.9.14_offline.sh"
+RUN wget "https://github.com/uxlfoundation/oneCCL/releases/download/2021.15.9/${ONECCL_INSTALLER}" && \
     bash "${ONECCL_INSTALLER}" -a --silent --eula accept && \
     rm "${ONECCL_INSTALLER}" && \
     echo "source /opt/intel/oneapi/setvars.sh --force" >> /root/.bashrc && \
@@ -164,7 +164,7 @@ RUN --mount=type=cache,target=/root/.cache/uv \
 # FIX triton
 RUN --mount=type=cache,target=/root/.cache/uv \
     uv pip uninstall triton triton-xpu && \
-    uv pip install triton-xpu==3.6.0
+    uv pip install triton-xpu==3.7.0
 
 # remove torch bundled oneccl to avoid conflicts
 RUN --mount=type=cache,target=/root/.cache/uv \

docker/docker-bake.hcl

@@ -20,7 +20,7 @@ variable "NVCC_THREADS" {
 }
 
 variable "TORCH_CUDA_ARCH_LIST" {
-  default = "8.0 8.9 9.0 10.0"
+  default = "8.0 8.9 9.0 10.0 11.0 12.0"
 }
 
 variable "COMMIT" {
@@ -88,7 +88,6 @@ target "test-ubuntu2404" {
   args = {
     UBUNTU_VERSION          = "24.04"
     GDRCOPY_OS_VERSION      = "Ubuntu24_04"
-    FLASHINFER_AOT_COMPILE  = "true"
   }
   output = ["type=docker"]
 }
@@ -100,7 +99,6 @@ target "openai-ubuntu2404" {
   args = {
     UBUNTU_VERSION          = "24.04"
     GDRCOPY_OS_VERSION      = "Ubuntu24_04"
-    FLASHINFER_AOT_COMPILE  = "true"
   }
   output = ["type=docker"]
 }

docker/versions.json

@@ -2,7 +2,7 @@
   "_comment": "Auto-generated from Dockerfile ARGs. Do not edit manually. Run: python tools/generate_versions_json.py",
   "variable": {
     "CUDA_VERSION": {
-      "default": "13.0.0"
+      "default": "13.0.2"
     },
     "PYTHON_VERSION": {
       "default": "3.12"
@@ -11,10 +11,10 @@
       "default": "22.04"
     },
     "BUILD_BASE_IMAGE": {
-      "default": "nvidia/cuda:13.0.0-devel-ubuntu22.04"
+      "default": "nvidia/cuda:13.0.2-devel-ubuntu22.04"
     },
     "FINAL_BASE_IMAGE": {
-      "default": "nvidia/cuda:13.0.0-base-ubuntu22.04"
+      "default": "nvidia/cuda:13.0.2-base-ubuntu22.04"
     },
     "GET_PIP_URL": {
       "default": "https://bootstrap.pypa.io/get-pip.py"
@@ -32,7 +32,7 @@
       "default": "false"
     },
     "TORCH_CUDA_ARCH_LIST": {
-      "default": "7.0 7.5 8.0 8.9 9.0 10.0 12.0"
+      "default": "7.5 8.0 8.6 8.9 9.0 10.0 11.0 12.0+PTX"
     },
     "MAX_JOBS": {
       "default": "2"
@@ -65,7 +65,7 @@
       "default": "true"
     },
     "FLASHINFER_VERSION": {
-      "default": "0.6.7"
+      "default": "0.6.8.post1"
     },
     "GDRCOPY_CUDA_VERSION": {
       "default": "12.8"

docs/benchmarking/cli.md

@@ -108,6 +108,38 @@ P99 ITL (ms):                            8.39
 ==================================================
 ```
 
+#### Results Visualization
+
+The `--plot-timeline` and `--plot-dataset-stats` can be used to generate respectively the requests completion timeline and dataset prompt and output tokens statistics, which can be useful for debugging purpose or for deeper analysis.
+
+```bash
+vllm bench serve \
+    --backend vllm \
+    --model meta-llama/Llama-3.1-8B-Instruct \
+    --endpoint /v1/completions \
+    --dataset-name sharegpt \
+    --dataset-path <your data path>/ShareGPT_V3_unfiltered_cleaned_split.json \
+    --num-prompts 100 \
+    --plot-timeline \
+    --timeline-itl-thresholds 2,5 \
+    --plot-dataset-stats \
+    --save-result
+```
+
+##### Interactive Timeline
+
+The generated timeline is an interactive visualization in the form of an HTML file that can be rendered in most browsers. To customize the ITL color thresholds, one can use `--timeline-itl-thresholds` flag (default: 25ms, 50ms)
+
+Example output:
+
+<iframe src="../../assets/contributing/vllm_bench_serve_timeline.html" width="100%" height="600" frameborder="0"></iframe>
+
+##### Dataset statistics
+
+The generated figure shows the input prompt and output tokens distribution.
+
+Example output: ![Dataset Statistics](../assets/contributing/vllm_bench_serve_dataset_stats.png)
+
 #### Custom Dataset
 
 If the dataset you want to benchmark is not supported yet in vLLM, even then you can benchmark on it using `CustomDataset`. Your data needs to be in `.jsonl` format and needs to have "prompt" field per entry, e.g., data.jsonl

docs/cli/README.md

@@ -163,7 +163,7 @@ Running with a local file:
 
 ```bash
 vllm run-batch \
-    -i offline_inference/openai_batch/openai_example_batch.jsonl \
+    -i features/openai_batch/openai_example_batch.jsonl \
     -o results.jsonl \
     --model meta-llama/Meta-Llama-3-8B-Instruct
 ```
@@ -172,7 +172,7 @@ Using remote file:
 
 ```bash
 vllm run-batch \
-    -i https://raw.githubusercontent.com/vllm-project/vllm/main/examples/offline_inference/openai_batch/openai_example_batch.jsonl \
+    -i https://raw.githubusercontent.com/vllm-project/vllm/main/examples/features/openai_batch/openai_example_batch.jsonl \
     -o results.jsonl \
     --model meta-llama/Meta-Llama-3-8B-Instruct
 ```

docs/configuration/conserving_memory.md

@@ -23,7 +23,7 @@ llm = LLM(model="ibm-granite/granite-3.1-8b-instruct", tensor_parallel_size=2)
 !!! note
     With tensor parallelism enabled, each process will read the whole model and split it into chunks, which makes the disk reading time even longer (proportional to the size of tensor parallelism).
 
-    You can convert the model checkpoint to a sharded checkpoint using [examples/offline_inference/save_sharded_state.py](../../examples/offline_inference/save_sharded_state.py). The conversion process might take some time, but later you can load the sharded checkpoint much faster. The model loading time should remain constant regardless of the size of tensor parallelism.
+    You can convert the model checkpoint to a sharded checkpoint using [examples/features/sharded_state/load_sharded_state_offline.py](../../examples/features/sharded_state/load_sharded_state_offline.py). The conversion process might take some time, but later you can load the sharded checkpoint much faster. The model loading time should remain constant regardless of the size of tensor parallelism.
 
 ## Quantization
 

docs/configuration/optimization.md

@@ -155,9 +155,9 @@ switch to `--physcpubind=<cpu-list> --membind=<node>`.
 
 These `--numa-bind*` options only apply to GPU execution processes. They do not
 configure the CPU backend's separate thread-affinity controls. Automatic
-GPU-to-NUMA detection is currently implemented for CUDA/NVML-based platforms;
-other GPU backends must provide explicit binding lists if they use these
-options.
+GPU-to-NUMA detection is currently implemented for CUDA/NVML-based as well as
+ROCM-based platforms; other GPU backends must provide explicit binding lists if
+they use these options.
 
 `--numa-bind-nodes` takes one non-negative NUMA node index per visible GPU, in
 the same order as the GPU indices.

docs/contributing/model/transcription.md

@@ -66,7 +66,7 @@ This is for controlling general behavior of the API when serving your model:
 
 See [Audio preprocessing and chunking](#audio-preprocessing-and-chunking) for what each field controls.
 
-Implement the prompt construction via [get_generation_prompt][vllm.model_executor.models.interfaces.SupportsTranscription.get_generation_prompt]. The server passes you the resampled waveform and task parameters; you return a valid [PromptType][vllm.inputs.llm.PromptType]. There are two common patterns:
+Implement the prompt construction via [get_generation_prompt][vllm.model_executor.models.interfaces.SupportsTranscription.get_generation_prompt]. The server builds a [SpeechToTextParams][vllm.config.speech_to_text.SpeechToTextParams] object that bundles the resampled waveform, task parameters, and request-specific options. Your model receives this single object and returns a valid [PromptType][vllm.inputs.llm.PromptType]. There are two common patterns:
 
 #### Multimodal LLM with audio embeddings (e.g., Voxtral, Gemma3n)
 
@@ -75,21 +75,20 @@ Return a dict containing `multi_modal_data` with the audio, and either a `prompt
 ??? code "get_generation_prompt()"
 
     ```python
+    from vllm.config.speech_to_text import SpeechToTextParams
+
     class YourASRModel(nn.Module, SupportsTranscription):
         ...
 
         @classmethod
         def get_generation_prompt(
             cls,
-            audio: np.ndarray,
-            stt_config: SpeechToTextConfig,
-            model_config: ModelConfig,
-            language: str | None,
-            task_type: Literal["transcribe", "translate"],
-            request_prompt: str,
-            to_language: str | None,
+            stt_params: SpeechToTextParams,
         ) -> PromptType:
-            # Example with a free-form instruction prompt
+            audio = stt_params.audio
+            stt_config = stt_params.stt_config
+            task_type = stt_params.task_type
+
             task_word = "Transcribe" if task_type == "transcribe" else "Translate"
             prompt = (
                 "<start_of_turn>user\n"
@@ -112,20 +111,22 @@ Return a dict with separate `encoder_prompt` and `decoder_prompt` entries:
 ??? code "get_generation_prompt()"
 
     ```python
+    from vllm.config.speech_to_text import SpeechToTextParams
+
     class YourASRModel(nn.Module, SupportsTranscription):
         ...
 
         @classmethod
         def get_generation_prompt(
             cls,
-            audio: np.ndarray,
-            stt_config: SpeechToTextConfig,
-            model_config: ModelConfig,
-            language: str | None,
-            task_type: Literal["transcribe", "translate"],
-            request_prompt: str,
-            to_language: str | None,
+            stt_params: SpeechToTextParams,
         ) -> PromptType:
+            audio = stt_params.audio
+            stt_config = stt_params.stt_config
+            language = stt_params.language
+            task_type = stt_params.task_type
+            request_prompt = stt_params.request_prompt
+
             if language is None:
                 raise ValueError("Language must be specified")
 
@@ -193,7 +194,7 @@ Provide a fast duration→token estimate to improve streaming usage statistics:
 
 The API server takes care of basic audio I/O and optional chunking before building prompts:
 
-- Resampling: Input audio is resampled to `SpeechToTextConfig.sample_rate` using `librosa`.
+- Resampling: Input audio is resampled to `SpeechToTextConfig.sample_rate` using `AudioResampler`.
 - Chunking: If `SpeechToTextConfig.allow_audio_chunking` is True and the duration exceeds `max_audio_clip_s`, the server splits the audio into overlapping chunks and generates a prompt per chunk. Overlap is controlled by `overlap_chunk_second`.
 - Energy-aware splitting: When `min_energy_split_window_size` is set, the server finds low-energy regions to minimize cutting within words.
 
@@ -206,22 +207,20 @@ Relevant server logic:
     async def _preprocess_speech_to_text(...):
         language = self.model_cls.validate_language(request.language)
         ...
-        y, sr = librosa.load(bytes_, sr=self.asr_config.sample_rate)
-        duration = librosa.get_duration(y=y, sr=sr)
+        y, sr = load_audio(bytes_, sr=self.asr_config.sample_rate)
+        duration = get_audio_duration(y=y, sr=sr)
         do_split_audio = (self.asr_config.allow_audio_chunking
                         and duration > self.asr_config.max_audio_clip_s)
         chunks = [y] if not do_split_audio else self._split_audio(y, int(sr))
         prompts = []
         for chunk in chunks:
-            prompt = self.model_cls.get_generation_prompt(
+            stt_params = request.build_stt_params(
                 audio=chunk,
                 stt_config=self.asr_config,
                 model_config=self.model_config,
-                language=language,
                 task_type=self.task_type,
-                request_prompt=request.prompt,
-                to_language=to_language,
             )
+            prompt = self.model_cls.get_generation_prompt(stt_params)
             prompts.append(prompt)
         return prompts, duration
     ```

docs/contributing/profiling.md

@@ -42,7 +42,7 @@ Traces can be visualized using <https://ui.perfetto.dev/>.
 
 #### Offline Inference
 
-Refer to [examples/offline_inference/simple_profiling.py](../../examples/offline_inference/simple_profiling.py) for an example.
+Refer to [examples/features/profiling/simple_profiling_offline.py](../../examples/features/profiling/simple_profiling_offline.py) for an example.
 
 #### OpenAI Server
 
@@ -206,8 +206,8 @@ Both the `vllm.utils.profiling.cprofile` and `vllm.utils.profiling.cprofile_cont
 used to profile a section of code.
 
 !!! note
-    The legacy import paths `vllm.utils.cprofile` and `vllm.utils.cprofile_context` are deprecated.
-    Please use `vllm.utils.profiling.cprofile` and `vllm.utils.profiling.cprofile_context` instead.
+    The `vllm.utils.profiling` helpers are deprecated and will be removed in
+    `v0.21`. Please use Python's `cProfile` module directly instead.
 
 ### Example usage - decorator
 

docs/deployment/k8s.md

@@ -4,6 +4,7 @@ Deploying vLLM on Kubernetes is a scalable and efficient way to serve machine le
 
 - [Deployment with CPUs](#deployment-with-cpus)
 - [Deployment with GPUs](#deployment-with-gpus)
+- [Serving with gRPC](#serving-with-grpc)
 - [Troubleshooting](#troubleshooting)
     - [Startup Probe or Readiness Probe Failure, container log contains "KeyboardInterrupt: terminated"](#startup-probe-or-readiness-probe-failure-container-log-contains-keyboardinterrupt-terminated)
 - [Conclusion](#conclusion)
@@ -387,6 +388,49 @@ INFO:     Uvicorn running on http://0.0.0.0:8000 (Press CTRL+C to quit)
 
       If the service is correctly deployed, you should receive a response from the vLLM model.
 
+## Serving with gRPC
+
+vLLM can serve models over gRPC instead of HTTP by passing the `--grpc` flag. This requires the optional gRPC dependencies:
+
+```bash
+pip install vllm[grpc]
+```
+
+When using `--grpc`, the server exposes the standard [gRPC Health Checking Protocol](https://github.com/grpc/grpc/blob/master/doc/health-checking.md) (`grpc.health.v1.Health`), which integrates with Kubernetes [native gRPC probes](https://kubernetes.io/docs/tasks/configure-pod-container/configure-liveness-readiness-startup-probes/#define-a-grpc-liveness-probe) (available since Kubernetes 1.24).
+
+To deploy with gRPC, change the `vllm serve` command to include `--grpc` and replace `httpGet` probes with `grpc` probes:
+
+```yaml
+containers:
+- name: mistral-7b
+  image: vllm/vllm-openai:latest
+  command: ["/bin/sh", "-c"]
+  args: [
+    "pip install vllm[grpc] && vllm serve mistralai/Mistral-7B-Instruct-v0.3 --grpc --port 50051 --trust-remote-code"
+  ]
+  ports:
+  - containerPort: 50051
+  livenessProbe:
+    grpc:
+      port: 50051
+    initialDelaySeconds: 120
+    periodSeconds: 10
+  readinessProbe:
+    grpc:
+      port: 50051
+    initialDelaySeconds: 120
+    periodSeconds: 5
+```
+
+!!! note
+    The gRPC health service checks the engine status on every probe. If the engine is unhealthy or the server is shutting down, the probe returns `NOT_SERVING`.
+
+You can also verify the health service manually with `grpcurl`:
+
+```bash
+grpcurl -plaintext localhost:50051 grpc.health.v1.Health/Check
+```
+
 ## Troubleshooting
 
 ### Startup Probe or Readiness Probe Failure, container log contains "KeyboardInterrupt: terminated"

docs/design/attention_backends.md

@@ -169,10 +169,10 @@ Priority is **1 = highest** (tried first).
 | ------- | ------- | ------ | --------- | ----------- | ---------- | ---- | --------- | --- | --------------- | ------------ |
 | `CPU_ATTN` | | fp16, bf16, fp32 | `auto` | Any | 32, 64, 80, 96, 112, 128, 160, 192, 224, 256, 512 | ❌ | ❌ | ❌ | All | N/A |
 | `FLASHINFER` | Native† | fp16, bf16 | `auto`, `float16`, `bfloat16`, `fp8`, `fp8_e4m3`, `fp8_e5m2` | 16, 32, 64 | 64, 128, 256 | ❌ | ❌ | ✅ | Decoder | 7.x-9.x |
-| `FLASHINFER` | TRTLLM† | fp16, bf16 | `auto`, `float16`, `bfloat16`, `fp8`, `fp8_e4m3`, `fp8_e5m2` | 16, 32, 64 | 64, 128, 256 | ✅ | ❌ | ✅ | Decoder | 10.0 |
+| `FLASHINFER` | TRTLLM† | fp16, bf16 | `auto`, `float16`, `bfloat16`, `fp8`, `fp8_e4m3`, `fp8_e5m2` | 16, 32, 64 | 64, 128, 256 | ✅ | ❌ | ✅ | Decoder | 10.x |
 | `FLASH_ATTN` | FA2* | fp16, bf16 | `auto`, `float16`, `bfloat16` | %16 | Any | ❌ | ❌ | ✅ | All | ≥8.0 |
 | `FLASH_ATTN` | FA3* | fp16, bf16 | `auto`, `float16`, `bfloat16`, `fp8`, `fp8_e4m3`, `fp8_e5m2` | %16 | Any | ✅ | ❌ | ✅ | All | 9.x |
-| `FLASH_ATTN` | FA4* | fp16, bf16 | `auto`, `float16`, `bfloat16` | %16 | Any | ❌ | ❌ | ✅ | All | ≥10.0 |
+| `FLASH_ATTN` | FA4* | fp16, bf16 | `auto`, `float16`, `bfloat16` | %16 | Any | ✅ | ❌ | ✅ | All | ≥10.0 |
 | `FLASH_ATTN_DIFFKV` | | fp16, bf16 | `auto` | Any | Any | ❌ | ❌ | ✅ | Decoder | Any |
 | `FLEX_ATTENTION` | | fp16, bf16, fp32 | `auto`, `float16`, `bfloat16` | Any | Any | ❌ | ✅ | ❌ | Decoder, Encoder Only | Any |
 | `ROCM_AITER_FA` | | fp16, bf16 | `auto`, `float16`, `bfloat16`, `fp8`, `fp8_e4m3`, `fp8_e5m2` | 16, 32 | 64, 128, 256 | ❌ | ❌ | ❌ | Decoder | N/A |
@@ -213,9 +213,9 @@ configuration.
 | `FLASHINFER_MLA` | fp16, bf16 | `auto`, `float16`, `bfloat16`, `fp8`, `fp8_e4m3` | 32, 64 | Any | ❌ | ❌ | ❌ | ❌ | Decoder | 10.x |
 | `FLASHINFER_MLA_SPARSE` | fp16, bf16 | `auto`, `float16`, `bfloat16`, `fp8`, `fp8_e4m3` | 32, 64 | 576 | ❌ | ✅ | ❌ | ❌ | Decoder | 10.x |
 | `FLASHMLA` | fp16, bf16 | `auto`, `float16`, `bfloat16`, `fp8`, `fp8_e4m3` | 64 | Any | ❌ | ❌ | ❌ | ✅ | Decoder | 9.x-10.x |
-| `FLASHMLA_SPARSE` | bf16 | `auto`, `bfloat16`, `fp8_ds_mla` | 64 | 576 | ❌ | ✅ | ❌ | ❌ | Decoder | 9.x-10.x |
+| `FLASHMLA_SPARSE` | bf16 | `auto`, `bfloat16`, `fp8_ds_mla` | 64 | 512, 576 | ❌ | ✅ | ❌ | ❌ | Decoder | 9.x-10.x |
 | `FLASH_ATTN_MLA` | fp16, bf16 | `auto`, `float16`, `bfloat16` | %16 | Any | ❌ | ❌ | ❌ | ✅ | Decoder | 9.x |
-| `ROCM_AITER_MLA` | fp16, bf16 | `auto`, `float16`, `bfloat16`, `fp8`, `fp8_e4m3`, `fp8_e5m2` | 1 | Any | ❌ | ❌ | ❌ | ❌ | Decoder | N/A |
+| `ROCM_AITER_MLA` | fp16, bf16 | `auto`, `float16`, `bfloat16`, `fp8`, `fp8_e4m3`, `fp8_e5m2` | %1 | Any | ❌ | ❌ | ❌ | ❌ | Decoder | N/A |
 | `ROCM_AITER_MLA_SPARSE` | fp16, bf16 | `auto`, `float16`, `bfloat16` | 1 | Any | ❌ | ✅ | ❌ | ❌ | Decoder | N/A |
 | `ROCM_AITER_TRITON_MLA` | fp16, bf16 | `auto` | Any | Any | ❌ | ❌ | ❌ | ❌ | Decoder | N/A |
 | `TRITON_MLA` | fp16, bf16 | `auto`, `float16`, `bfloat16`, `fp8`, `fp8_e4m3` | %16 | Any | ❌ | ❌ | ❌ | ✅ | Decoder | Any |

docs/design/cuda_graphs_multimodal.md

@@ -52,14 +52,14 @@ For each graph replay:
 
 When `mm_encoder_tp_mode="data"`, the manager distributes images across TP ranks using load-balanced assignment via `get_load_balance_assignment`, executes locally on each rank, then gathers results back in the original order via `tensor_model_parallel_all_gather`.
 
-### Video inference support (experimental)
+### Video inference support
 
 Following <https://github.com/vllm-project/vllm/pull/35963> (ViT full CUDA graph support for image inference), <https://github.com/vllm-project/vllm/pull/38061> extends the encoder CUDA graph framework to support video inference for Qwen3-VL. Previously, the CUDA graph capture/replay path only handled image inputs (`pixel_values` + `image_grid_thw`). Video inputs use different keys (`pixel_values_videos` + `video_grid_thw`) and require larger `cu_seqlens` buffers because each video item contributes multiple frames (`T` attention sequences). This PR generalizes the protocol and manager to handle both modalities through a single shared graph manager.
 
 !!! note
     Video CUDA graphs are automatically disabled when EVS (Efficient Video Sampling) pruning is enabled, since EVS makes the token count data-dependent and incompatible with CUDA graph capture.
 
-    Currently, we only support image-only or video-only inputs when enabling CUDA graph, mixed inputs (image + video) are not supported yet (we will work on it in the near future). Thus, it's recommended to turn off the image modality by `--limit-mm-per-prompt '{"image": 0}'` for video-only inputs.
+    Mixed inputs (image+video) per prompt are also supported now.
 
 ## Model integration via `SupportsEncoderCudaGraph`
 
@@ -76,6 +76,7 @@ Models opt-in to encoder CUDA Graphs by implementing the [SupportsEncoderCudaGra
 * `encoder_cudagraph_forward(...)` — forward pass using precomputed buffers (called during capture and replay).
 * `encoder_eager_forward(...)` — fallback eager forward when no graph fits.
 * `get_input_modality(...)` - return the modality of the inputs.
+* `get_max_frames_per_video()` - return model-specific max frames per video.
 
 !!! note
     The `SupportsEncoderCudaGraph` protocol is designed to be model-agnostic. New vision encoder models can opt-in by implementing the protocol methods without modifying the manager.
@@ -96,7 +97,7 @@ Three fields in `CompilationConfig` control encoder CUDA Graphs:
 * `cudagraph_mm_encoder` (`bool`, default `False`) — enable CUDA Graph capture for multimodal encoder. When enabled, captures the full encoder forward as a CUDA Graph for each token budget level.
 * `encoder_cudagraph_token_budgets` (`list[int]`, default `[]`) — token budget levels for capture. If empty (default), auto-inferred from model architecture as power-of-2 levels. User-provided values override auto-inference.
 * `encoder_cudagraph_max_vision_items_per_batch` (`int`, default `0`) — maximum number of images/videos per batch during capture. If 0 (default), auto-inferred as `max_budget // min_budget`.
-* `encoder_cudagraph_max_frames_per_batch` (`int`, default `0`) — maximum number of video frames per batch during capture. If 0 (default), auto-inferred as `encoder_cudagraph_max_vision_items_per_batch * 2` (to be optimized).
+* `encoder_cudagraph_max_frames_per_batch` (`int`, default `None`) — maximum number of video frames per batch during capture. If `None` (default), auto-inferred as `encoder_cudagraph_max_vision_items_per_batch * max_frames_per_video` (`max_frames_per_video` is a model-specific value according to its `processing_info`). If we limit the video count per prompt to `0`, it will also be set to `0` (i.e., fall back to image-only mode).
 
 ## Usage guide
 
@@ -142,7 +143,6 @@ Enable encoder CUDA Graphs via `compilation_config`:
 
 ```bash
 vllm serve Qwen/Qwen3-VL-32B \
-  --limit-mm-per-prompt '{"image": 0}' \
   --compilation-config '{"cudagraph_mm_encoder": true}'
 ```
 
@@ -150,7 +150,6 @@ With explicit budgets:
 
 ```bash
 vllm serve Qwen/Qwen3-VL-32B \
-  --limit-mm-per-prompt '{"image": 0}' \
   --compilation-config '{"cudagraph_mm_encoder": true, "encoder_cudagraph_token_budgets": [2048, 4096, 8192, 13824], "encoder_cudagraph_max_vision_items_per_batch": 8, "encoder_cudagraph_max_frames_per_batch": 64}'
 ```
 
@@ -169,7 +168,6 @@ compilation_config = {
 
 model = vllm.LLM(
     model="Qwen/Qwen3-VL-32B",
-    limit_mm_per_prompt='{"image": 0}',
     compilation_config=compilation_config,
 )
 ```

docs/design/fusions.md

@@ -21,6 +21,7 @@ or just on the low or high end.
 | Fusion                                                                         | `PassConfig` flag            | Fused operations                               | Default at                     | E2E Speedup        | Fullgraph | `num_tokens` |
 | ------------------------------------------------------------------------------ | ---------------------------- | ---------------------------------------------- | ------------------------------ | ------------------ | --------- | ------------ |
 | [AllReduce + RMSNorm](#allreduce--rmsnorm-fuse_allreduce_rms)                  | `fuse_allreduce_rms`         | All-reduce → RMSNorm (+residual_add) (→ quant) | O2 (Hopper/Blackwell + TP > 1) | 5-20%              | No        | Low          |
+| [MiniMax QK Norm](#minimax-qk-norm-fuse_minimax_qk_norm)                       | `fuse_minimax_qk_norm`       | Q/K variance all-reduce → Q/K RMSNorm          | Off by default                 | 2-3%               | No        | Low          |
 | [Attention + Quant](#attention--quantization-fuse_attn_quant)                  | `fuse_attn_quant`            | Attention output → FP8/NVFP4 quant             | Off by default                 | 3-7%               | Yes       | Always       |
 | [MLA Attention + Quant](#attention--quantization-fuse_attn_quant)              | `fuse_attn_quant`            | MLA Attention output → FP8/NVFP4 quant         | Off by default                 | TBD                | Yes       | Always       |
 | [RoPE + KV-Cache Update](#rope--kv-cache-update-fuse_rope_kvcache)             | `fuse_rope_kvcache`          | Rotary embedding → KV cache write              | O2 (ROCm/AITER only)           | 2-4%               | No        | Low          |
@@ -30,6 +31,7 @@ or just on the low or high end.
 | [RMSNorm + Quant](#rmsnorm--quantization-fuse_norm_quant)                      | `fuse_norm_quant`            | RMSNorm (+residual add) → FP8/FP4 quant        | O1 (conditional)               | 1-4%               | No        | Always       |
 | [SiLU+Mul + Quant](#silumul--quantization-fuse_act_quant)                      | `fuse_act_quant`             | SiLU+Mul activation → FP8/FP4 quant            | O1 (conditional)               | 1-4%               | No        | Always       |
 | [RMSNorm + Padding](#rmsnorm--padding-fuse_act_padding)                        | `fuse_act_padding`           | Residual add + RMSNorm → padding               | O1 (ROCm/AITER only)           | TBD                | No        | Always       |
+| [MLA Dual RMSNorm](#mla-dual-rmsnorm-fuse_mla_dual_rms_norm)                   | `fuse_mla_dual_rms_norm`     | Paired Q + KV RMSNorm → single kernel          | O1 (ROCm/AITER only)           | ~2%                | No        | Always       |
 
 ## Support Matrix
 
@@ -40,8 +42,9 @@ The table below lists the quantization schemes supported by each fusion on each
 | Fusion                       | SM100 (Blackwell)                        | SM90 (Hopper)                            | SM89 (Ada)                               | SM80 (Ampere) | ROCm                                     |
 | ---------------------------- | ---------------------------------------- | ---------------------------------------- | ---------------------------------------- | ------------- | ---------------------------------------- |
 | `fuse_allreduce_rms`         | FP16/BF16, FP8 static, NVFP4             | FP16/BF16, FP8 static                    | —                                        | —             | —                                        |
+| `fuse_minimax_qk_norm`\*     | FP16/BF16                                | FP16/BF16                                | FP16/BF16                                | FP16/BF16     | —                                        |
 | `fuse_attn_quant`\*          | FP8 static\*, NVFP4\*                    | FP8 static\*                             | FP8 static\*                             | —             | FP8 static\*                             |
-| `fuse_attn_quant` (MLA)\*    | FP8 static\*, NVFP4\*                    | FP8 static\*                             | FP8 static\*                             | —             | FP8 static(untested)\*                   |
+| `fuse_attn_quant` (MLA)\*    | FP8 static\*, FP8 per-group\*, NVFP4\*   | FP8 static\*, FP8 per-group\*            | FP8 static\*, FP8 per-group\*            | —             | FP8 static\* (untested)                  |
 | `fuse_rope_kvcache`          | —                                        | —                                        | —                                        | —             | FP16/BF16                                |
 | `enable_qk_norm_rope_fusion` | FP16/BF16                                | FP16/BF16                                | FP16/BF16†                               | FP16/BF16†    | —                                        |
 | `enable_sp`                  | FP16/BF16, FP8 static†                   | FP16/BF16, FP8 static                    | FP16/BF16†                               | FP16/BF16†    | —                                        |
@@ -49,11 +52,15 @@ The table below lists the quantization schemes supported by each fusion on each
 | `fuse_norm_quant`            | FP8 static, FP8 per-token, FP8 per-group | FP8 static, FP8 per-token, FP8 per-group | FP8 static, FP8 per-token, FP8 per-group | —             | FP8 static, FP8 per-token, FP8 per-group |
 | `fuse_act_quant`             | FP8 static, NVFP4                        | FP8 static, FP8 per-group (128/64)       | FP8 static, FP8 per-group (128/64)       | —             | FP8 per-group                            |
 | `fuse_act_padding`           | —                                        | —                                        | —                                        | —             | FP16/BF16                                |
+| `fuse_mla_dual_rms_norm`     | —                                        | —                                        | —                                        | —             | BF16                                     |
 
 \* `fuse_attn_quant` support depends on the attention backend in use; not all backends support
 fused quantization output. See the [`fuse_attn_quant` section](#attention--quantization-fuse_attn_quant)
 for per-backend details.
 
+\* `fuse_minimax_qk_norm` is a model-specific pass for `MiniMaxM2ForCausalLM`. It also requires
+tensor parallelism (`tp_size > 1`) and the CUDA custom op `minimax_allreduce_rms_qk`.
+
 † `enable_sp` and `fuse_gemm_comms` are only autoconfigured for SM90 today;
 other architectures support requires setting `PassConfig.sp_min_token_num` explicitly.
 SM100 support also requires setting `VLLM_DISABLED_KERNELS=FlashInferFP8ScaledMMLinearKernel`.
@@ -145,7 +152,7 @@ standard `Attention` and `MLAAttention` (used by DeepSeek-V2/V3/R1 models). Patt
 
 - `FLASHINFER`: CUDA sm100+ with FlashInfer installed
 
-`MLAAttention → FP8 static quant` / `MLAAttention → NVFP4 dynamic quant`:
+`MLAAttention → FP8 static, FP8 per-group, NVFP4 dynamic quant`
 
 The MLA fusion operates at the graph level on the `unified_mla_attention_with_output` op and works
 with all MLA decode and prefill backend combinations. Unlike standard `Attention` backends (where
@@ -184,6 +191,35 @@ If these conditions are set, the fusion is enabled automatically for optimizatio
 
 - Pass: [`vllm/compilation/passes/fusion/rope_kvcache_fusion.py`](https://github.com/vllm-project/vllm/blob/main/vllm/compilation/passes/fusion/rope_kvcache_fusion.py)
 
+### MiniMax QK Norm (`fuse_minimax_qk_norm`)
+
+!!! info
+    This is a MiniMax-specific compile pass. It is currently only enabled when all of the following hold:
+    the model architecture is `MiniMaxM2ForCausalLM`, tensor parallelism is enabled (`tp_size > 1`),
+    and the CUDA custom op `minimax_allreduce_rms_qk` is available. It is not enabled by default at any
+    optimization level.
+
+**What it fuses.** Fuses the MiniMax M2 Q/K normalization path that performs an all-reduce over the
+per-token Q/K variances before applying RMS normalization to Q and K.
+
+This pass is distinct from [`enable_qk_norm_rope_fusion`](#qk-norm--rope-enable_qk_norm_rope_fusion):
+`fuse_minimax_qk_norm` targets MiniMax M2's tensor-parallel all-reduce + RMSNorm sequence, while
+`enable_qk_norm_rope_fusion` targets the later Q/K RMSNorm + RoPE sequence used by several other models.
+
+Example:
+
+```bash
+vllm serve MiniMaxAI/MiniMax-M2.5 \
+  --tensor-parallel-size 4 \
+  --compilation-config '{"mode": 3, "pass_config": {"fuse_minimax_qk_norm": true}}'
+```
+
+**Code locations.**
+
+- Pass: [`vllm/compilation/passes/fusion/minimax_qk_norm_fusion.py`](https://github.com/vllm-project/vllm/blob/main/vllm/compilation/passes/fusion/minimax_qk_norm_fusion.py)
+- CUDA op: [`csrc/minimax_reduce_rms_kernel.cu`](https://github.com/vllm-project/vllm/blob/main/csrc/minimax_reduce_rms_kernel.cu) (`minimax_allreduce_rms_qk`)
+- Workspace helper: [`vllm/model_executor/layers/mamba/lamport_workspace.py`](https://github.com/vllm-project/vllm/blob/main/vllm/model_executor/layers/mamba/lamport_workspace.py)
+
 ### Sequence Parallelism (`enable_sp`)
 
 **What it fuses.** Replaces all-reduce collectives with reduce-scatter + local RMSNorm + all-gather,
@@ -347,6 +383,44 @@ when the hidden size is 2880 and AITER Triton GEMMs *not* enabled.
 
 - Pass: [`vllm/compilation/passes/fusion/rocm_aiter_fusion.py`](https://github.com/vllm-project/vllm/blob/main/vllm/compilation/passes/fusion/rocm_aiter_fusion.py) (`RocmAiterTritonAddRMSNormPadFusionPass`)
 
+### MLA Dual RMSNorm (`fuse_mla_dual_rms_norm`)
+
+!!! info
+    ROCm/AITER-only. Targeted at DeepSeek-V3 / Kimi-K2 MLA attention.
+
+!!! note
+    When the native implementation of `rms_norm` is used (the default on CUDA and
+    ROCm for now), Inductor's built-in fusion already handles merging these norms
+    automatically. This explicit pass targets the case where AITER's custom
+    `rms_norm` op is active, which Inductor cannot fuse on its own.
+
+**What it fuses.** Fuses the paired `q_a_layernorm` and `kv_a_layernorm` RMS norm
+operations in MLA attention into a single `fused_qk_rmsnorm` HIP kernel call via AITER,
+reducing kernel launch overhead from 2 launches to 1 per MLA layer.
+
+```text
+# Unfused:
+q_c, kv_lora = split(projected, [q_dim, kv_dim])
+kv_c, k_pe   = split(kv_lora,  [kv_c_dim, k_pe_dim])
+q_c  = rms_norm(q_c,  q_weight,  eps)
+kv_c = rms_norm(kv_c, kv_weight, eps)
+
+# Fused:
+q_c, kv_lora = split(projected, [q_dim, kv_dim])
+kv_c, k_pe   = split(kv_lora,  [kv_c_dim, k_pe_dim])
+q_normed, kv_normed = fused_mla_dual_rms_norm(
+    q_c, q_weight, kv_c, kv_weight, eps1, eps2)
+```
+
+Requires: AMD ROCm with AITER enabled. Enabled by default at optimization level O1 and above
+when AITER is available.
+
+**Code locations.**
+
+- Pass: [`vllm/compilation/passes/fusion/rocm_aiter_fusion.py`](https://github.com/vllm-project/vllm/blob/main/vllm/compilation/passes/fusion/rocm_aiter_fusion.py) (`MLADualRMSNormFusionPass`)
+- Custom op: [`vllm/_aiter_ops.py`](https://github.com/vllm-project/vllm/blob/main/vllm/_aiter_ops.py) (`fused_mla_dual_rms_norm`)
+- AITER kernel: [`fused_qk_rmsnorm`](https://github.com/ROCm/aiter/pull/2442)
+
 ## See Also
 
 - [Optimization Levels](optimization_levels.md) — high-level presets that set

docs/design/moe_kernel_features.md

@@ -36,7 +36,7 @@ th {
 | deepep_high_throughput | standard | fp8 | G(128),A,T<sup>2</sup> | Y | Y | [`DeepEPHTPrepareAndFinalize`][vllm.model_executor.layers.fused_moe.prepare_finalize.deepep_ht.DeepEPHTPrepareAndFinalize] |
 | deepep_low_latency | batched | fp8 | G(128),A,T<sup>3</sup> | Y | Y | [`DeepEPLLPrepareAndFinalize`][vllm.model_executor.layers.fused_moe.prepare_finalize.deepep_ll.DeepEPLLPrepareAndFinalize] |
 | flashinfer_nvlink_two_sided | standard | nvfp4,fp8 | G,A,T | N | N | [`FlashInferNVLinkTwoSidedPrepareAndFinalize`][vllm.model_executor.layers.fused_moe.prepare_finalize.flashinfer_nvlink_two_sided.FlashInferNVLinkTwoSidedPrepareAndFinalize] |
-| flashinfer_nvlink_one_sided | standard | nvfp4 | G,A,T | N | N | [`FlashInferNVLinkOneSidedPrepareAndFinalize`][vllm.model_executor.layers.fused_moe.prepare_finalize.flashinfer_nvlink_one_sided.FlashInferNVLinkOneSidedPrepareAndFinalize] |
+| flashinfer_nvlink_one_sided | standard | nvfp4,bf16,mxfp8 | G,A,T | N | N | [`FlashInferNVLinkOneSidedPrepareAndFinalize`][vllm.model_executor.layers.fused_moe.prepare_finalize.flashinfer_nvlink_one_sided.FlashInferNVLinkOneSidedPrepareAndFinalize] |
 
 !!! info "Table key"
     1. All types: mxfp4, nvfp4, int4, int8, fp8
@@ -83,8 +83,8 @@ To be used with a particular `FusedMoEPrepareAndFinalizeModular` subclass, MoE k
 | triton | standard | all<sup>1</sup> | G,A,T | silu, gelu,</br>swigluoai,</br>silu_no_mul,</br>gelu_no_mul | Y | Y | [`fused_experts`][vllm.model_executor.layers.fused_moe.fused_moe.fused_experts],</br>[`TritonExperts`][vllm.model_executor.layers.fused_moe.fused_moe.TritonExperts] |
 | triton (batched) | batched | all<sup>1</sup> | G,A,T | silu, gelu | <sup>6</sup> | Y | [`BatchedTritonExperts`][vllm.model_executor.layers.fused_moe.fused_batched_moe.BatchedTritonExperts] |
 | deep gemm | standard,</br>batched | fp8 | G(128),A,T | silu, gelu | <sup>6</sup> | Y | </br>[`DeepGemmExperts`][vllm.model_executor.layers.fused_moe.experts.deep_gemm_moe.DeepGemmExperts],</br>[`BatchedDeepGemmExperts`][vllm.model_executor.layers.fused_moe.experts.batched_deep_gemm_moe.BatchedDeepGemmExperts] |
-| cutlass_fp4 | standard,</br>batched | nvfp4 | A,T | silu | Y | Y | [`CutlassExpertsFp4`][vllm.model_executor.layers.fused_moe.cutlass_moe.CutlassExpertsFp4] |
-| cutlass_fp8 | standard,</br>batched | fp8 | A,T | silu, gelu | Y | Y | [`CutlassExpertsFp8`][vllm.model_executor.layers.fused_moe.cutlass_moe.CutlassExpertsFp8],</br>[`CutlasBatchedExpertsFp8`][vllm.model_executor.layers.fused_moe.cutlass_moe.CutlassBatchedExpertsFp8] |
+| cutlass_fp4 | standard,</br>batched | nvfp4 | A,T | silu | Y | Y | [`CutlassExpertsFp4`][vllm.model_executor.layers.fused_moe.experts.cutlass_moe.CutlassExpertsFp4] |
+| cutlass_fp8 | standard,</br>batched | fp8 | A,T | silu, gelu | Y | Y | [`CutlassExpertsFp8`][vllm.model_executor.layers.fused_moe.experts.cutlass_moe.CutlassExpertsFp8],</br>[`CutlasBatchedExpertsFp8`][vllm.model_executor.layers.fused_moe.experts.cutlass_moe.CutlassBatchedExpertsFp8] |
 | flashinfer | standard | nvfp4,</br>fp8 | T | <sup>5</sup> | N | Y | [`FlashInferExperts`][vllm.model_executor.layers.fused_moe.flashinfer_cutlass_moe.FlashInferExperts] |
 | gpt oss triton | standard | N/A | N/A | <sup>5</sup> | Y | Y | [`triton_kernel_fused_experts`][vllm.model_executor.layers.fused_moe.experts.gpt_oss_triton_kernels_moe.triton_kernel_fused_experts],</br>[`OAITritonExperts`][vllm.model_executor.layers.fused_moe.experts.gpt_oss_triton_kernels_moe.OAITritonExperts] |
 | marlin | standard,</br>batched | <sup>3</sup> / N/A | <sup>3</sup> / N/A | silu,</br>swigluoai | Y | Y | [`fused_marlin_moe`][vllm.model_executor.layers.fused_moe.fused_marlin_moe.fused_marlin_moe],</br>[`MarlinExperts`][vllm.model_executor.layers.fused_moe.fused_marlin_moe.MarlinExperts],</br>[`BatchedMarlinExperts`][vllm.model_executor.layers.fused_moe.fused_marlin_moe.BatchedMarlinExperts] |

docs/design/optimization_levels.md

@@ -56,6 +56,7 @@ Fusions:
 - `-cc.pass_config.fuse_norm_quant=True`*
 - `-cc.pass_config.fuse_act_quant=True`*
 - `-cc.pass_config.fuse_act_padding=True`†
+- `-cc.pass_config.fuse_mla_dual_rms_norm=True`†
 
 \* These fusions are only enabled when either op is using a custom kernel, otherwise Inductor fusion is better.</br>
 † These fusions are ROCm-only and require AITER.

docs/features/automatic_prefix_caching.md

@@ -11,7 +11,7 @@ Automatic Prefix Caching (APC in short) caches the KV cache of existing queries,
 
 Set `enable_prefix_caching=True` in vLLM engine to enable APC. Here is an example:
 
-[examples/offline_inference/automatic_prefix_caching.py](../../examples/offline_inference/automatic_prefix_caching.py)
+[examples/features/automatic_prefix_caching/automatic_prefix_caching_offline.py](../../examples/features/automatic_prefix_caching/automatic_prefix_caching_offline.py)
 
 ## Example workloads
 

docs/features/batch_invariance.md

@@ -104,7 +104,7 @@ for output in outputs:
 Batch invariance has been tested and verified on the following models:
 
 - **DeepSeek series**: `deepseek-ai/DeepSeek-V3`, `deepseek-ai/DeepSeek-V3-0324`, `deepseek-ai/DeepSeek-R1`, `deepseek-ai/DeepSeek-V3.1`
-- **Qwen3 (Dense)**: `Qwen/Qwen3-1.7B`, `Qwen/Qwen3-8B`
+- **Qwen3 (Dense)**: `Qwen/Qwen3-1.7B`, `Qwen/Qwen3-8B`, `Qwen/Qwen3-4B-AWQ`, `Qwen/Qwen3-8B-AWQ`
 - **Qwen3 (MoE)**: `Qwen/Qwen3-30B-A3B`, `Qwen/Qwen3-Next-80B-A3B-Instruct`
 - **Qwen2.5**: `Qwen/Qwen2.5-0.5B-Instruct`, `Qwen/Qwen2.5-1.5B-Instruct`, `Qwen/Qwen2.5-3B-Instruct`, `Qwen/Qwen2.5-7B-Instruct`, `Qwen/Qwen2.5-14B-Instruct`, `Qwen/Qwen2.5-32B-Instruct`
 - **Llama 3**: `meta-llama/Llama-3.1-8B-Instruct`, `meta-llama/Llama-3.2-1B-Instruct`

docs/features/context_extension.md

@@ -0,0 +1,70 @@
+# Context Extension
+
+!!! note
+    The `--rope-scaling` parameter used in older versions of vLLM is no longer supported. Please use the `--hf-overrides` method with `rope_parameters` instead.
+This directory contains examples for extending the context length of models using vLLM.
+
+## Offline Inference Example
+
+The [`context_extension.py`](../../examples/features/context_extension/context_extension_offline.py) script demonstrates how to extend the context length of a Qwen model using the YARN method (rope_parameters) and run a simple chat example.
+
+### Usage
+
+```bash
+python examples/features/context_extension/context_extension_offline.py
+```
+
+## OpenAI Online Method
+
+You can also use vLLM's OpenAI-compatible API to serve models with extended context length.
+
+### Usage
+
+Run the vLLM server with the following command to extend the context length using YARN:
+
+```bash
+vllm serve Qwen/Qwen3-0.6B \
+  --hf-overrides '{"rope_parameters": {"factor": 4.0, "original_max_position_embeddings": 32768, "rope_theta": 1000000, "rope_type": "yarn"}}' \
+  --max-model-len 131072
+```
+
+### Client Example
+
+After starting the server, you can use the OpenAI Python client to interact with it:
+
+```python
+from openai import OpenAI
+
+client = OpenAI(
+    base_url="http://localhost:8000/v1",
+    api_key="token-abc123"  # Dummy API key, required by the client
+)
+
+response = client.chat.completions.create(
+    model="Qwen/Qwen3-0.6B",
+    messages=[
+        {"role": "system", "content": "You are a helpful assistant"},
+        {"role": "user", "content": "Hello"}
+    ],
+    max_tokens=128,
+    temperature=0.8,
+    top_p=0.95
+)
+
+print(response.choices[0].message.content)
+```
+
+### Key Parameters
+
+The available parameters depend on the `rope_type` you choose. For detailed information about all supported RoPE types and their specific parameters, please refer to the [Hugging Face Transformers RoPE documentation](https://huggingface.co/docs/transformers/main/en/internal/rope_utils#transformers.RopeParameters).
+
+Common parameters include:
+
+- `rope_type`: The type of RoPE implementation (e.g., "yarn", "linear", "dynamic")
+- `factor`: The factor by which to extend the context length
+- `original_max_position_embeddings`: The original maximum position embeddings of the model
+
+The following parameters are specific to vLLM:
+
+- `max_model_len`: The new maximum sequence length after extension (original * factor).
+  Used for KV cache pre‑allocation and request limit at serving time.

docs/features/lora.md

@@ -47,7 +47,7 @@ the third parameter is the path to the LoRA adapter.
     )
     ```
 
-Check out [examples/offline_inference/multilora_inference.py](../../examples/offline_inference/multilora_inference.py) for an example of how to use LoRA adapters with the async engine and how to use more advanced configuration options.
+Check out [examples/features/lora/multilora_offline.py](../../examples/features/lora/multilora_offline.py) for an example of how to use LoRA adapters with the async engine and how to use more advanced configuration options.
 
 ## Serving LoRA Adapters