forked from Karylab-cklius/vllm
cb7eb5c9b4
Move the MiniMax M3 sparse-attention backend and Triton kernels out of the
shared attention tree into the model definition (mirroring deepseek_v4), split
the kernels by role, and tighten the backend, then port the dedicated split-K
decode top-k from the sglang reference.
Structure:
- vllm/v1/attention/backends/minimax_m3_sparse.py
-> vllm/models/minimax_m3/common/sparse_attention.py
- vllm/v1/attention/ops/minimax_m3_sparse_ops.py split into
common/ops/index_topk.py (index-score + top-k kernels)
common/ops/sparse_attn.py (block-sparse GQA attention kernels)
Pure-Triton/cross-platform, so under common/ (not nvidia/).
- registry enum + doc generator RELEVANT_PATTERNS repointed; the auto-generated
attention_backends.md now has a dedicated "MiniMax M3 Sparse Attention" section.
Backend cleanups:
- Drop redundant metadata fields (sparse-selection params, decode max_query_len,
decode cu_seqlens_q/context_lens) and read them from the layer/impl instead.
- Impl ctor takes explicit named args instead of kwargs.get lookups.
- Cudagraph support UNIFORM_SINGLE_TOKEN_DECODE -> UNIFORM_BATCH; reorder
threshold and decode buffer scale by 1 + num_speculative_tokens.
- Unify the per-request context-length buffer (max_num_batched_tokens).
- Declare bf16-only KV cache.
Decode top-k (ported from sglang minimax_sparse_ops/decode):
- Split-K top-k: per-chunk partial top-k (_topk_index_partial_kernel) + merge
(_topk_index_merge_kernel), replacing the single-program prefill top-k reuse.
- init/local block forcing moved into the decode score kernel (matches sglang).
- Verified against a torch reference (selected-block sets match across varied
seq lengths, with/without init+local forcing, and short sequences).
Signed-off-by: Yongye Zhu <zyy1102000@gmail.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
hide
| hide | ||
|---|---|---|
|
Welcome to vLLM
Easy, fast, and cheap LLM serving for everyone
<script async defer src="https://buttons.github.io/buttons.js"></script> Star Watch Fork
vLLM is a fast and easy-to-use library for LLM inference and serving.
Originally developed in the Sky Computing Lab at UC Berkeley, vLLM has grown into one of the most active open-source AI projects built and maintained by a diverse community of many dozens of academic institutions and companies from over 2000 contributors.
Where to get started with vLLM depends on the type of user. If you are looking to:
- Run open-source models on vLLM, we recommend starting with the Quickstart Guide
- Build applications with vLLM, we recommend starting with the User Guide
- Build vLLM, we recommend starting with Developer Guide
For information about the development of vLLM, see:
vLLM is fast with:
- State-of-the-art serving throughput
- Efficient management of attention key and value memory with PagedAttention
- Continuous batching of incoming requests, chunked prefill, prefix caching
- Fast and flexible model execution with piecewise and full CUDA/HIP graphs
- Quantization: FP8, MXFP8/MXFP4, NVFP4, INT8, INT4, GPTQ/AWQ, GGUF, compressed-tensors, ModelOpt, TorchAO, and more
- Optimized attention kernels including FlashAttention, FlashInfer, TRTLLM-GEN, FlashMLA, and Triton
- Optimized GEMM/MoE kernels for various precisions using CUTLASS, TRTLLM-GEN, CuTeDSL
- Speculative decoding including n-gram, suffix, EAGLE, DFlash
- Automatic kernel generation and graph-level transformations using torch.compile
- Disaggregated prefill, decode, and encode
vLLM is flexible and easy to use with:
- Seamless integration with popular Hugging Face models
- High-throughput serving with various decoding algorithms, including parallel sampling, beam search, and more
- Tensor, pipeline, data, expert, and context parallelism for distributed inference
- Streaming outputs
- Generation of structured outputs using xgrammar or guidance
- Tool calling and reasoning parsers
- OpenAI-compatible API server, plus Anthropic Messages API and gRPC support
- 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:
- Decoder-only LLMs (e.g., Llama, Qwen, Gemma)
- Mixture-of-Expert LLMs (e.g., Mixtral, DeepSeek-V3, Qwen-MoE, GPT-OSS)
- Hybrid attention and state-space models (e.g., Mamba, Qwen3.5)
- Multi-modal models (e.g., LLaVA, Qwen-VL, Pixtral)
- Embedding and retrieval models (e.g., E5-Mistral, GTE, ColBERT)
- Reward and classification models (e.g., Qwen-Math)
Find the full list of supported models here.
For more information, check out the following:
- vLLM announcing blog post (intro to PagedAttention)
- vLLM paper (SOSP 2023)
- How continuous batching enables 23x throughput in LLM inference while reducing p50 latency by Cade Daniel et al.
- vLLM Meetups