[ROCm][CI] Remove mxfp4 test skips after amd-quark 0.12 release (#47330)

Signed-off-by: Micah Williamson <micah.williamson@amd.com>
Signed-off-by: Andreas Karatzas <akaratza@amd.com>
Signed-off-by: Andreas Karatzas <Andreas.Karatzas@amd.com>
Co-authored-by: Andreas Karatzas <Andreas.Karatzas@amd.com>
Co-authored-by: fxmarty-amd <felmarty@amd.com>
Co-authored-by: Andreas Karatzas <akaratza@amd.com>

requirements/rocm.txt

@@ -21,7 +21,7 @@ conch-triton-kernels==1.2.1
 timm>=1.0.17
 # amd-quark: required for Quark quantization on ROCm
 # To be consistent with test_quark.py
-amd-quark>=0.8.99
+amd-quark==0.12.post1
 tilelang==0.1.10
 # Required apache-tvm-ffi matching tilelang version
 apache-tvm-ffi==0.1.10 

tests/evals/gsm8k/test_gsm8k_correctness.py

@@ -9,30 +9,16 @@ pytest -s -v tests/evals/gsm8k/test_gsm8k_correctness.py \
     --config-list-file=configs/models-small.txt
 """
 
-import importlib.metadata
 import shlex
-from importlib.util import find_spec
 
 import pytest
-import torch
 import yaml
-from packaging import version
 
 from tests.utils import RemoteOpenAIServer
 from vllm.platforms import current_platform
 
 from .gsm8k_eval import evaluate_gsm8k
 
-# MXFP4 via quark requires amd-quark >= 0.12 on torch >= 2.11.
-# Earlier torch releases work with older quark versions. See
-# https://github.com/amd/Quark/issues/34
-# TODO: Remove once amd-quark>=0.12.0
-QUARK_MXFP4_TORCH_COMPATIBLE = find_spec("quark") is not None and (
-    version.parse(importlib.metadata.version("amd-quark")) >= version.parse("0.12.0")
-    if version.parse(torch.__version__.split("+")[0]) >= version.parse("2.11")
-    else True
-)
-
 DEFAULT_STARTUP_MAX_WAIT_SECONDS = 1200
 
 
@@ -109,19 +95,16 @@ def test_gsm8k_correctness(config_filename):
             "Skipping DeepSeek-V3.2 and DeepSeek-R1 on ROCm platforms "
             "due to agent pool disk space issues and pod evictions."
         )
-    if current_platform.is_rocm() and ("Qwen3.5-35B-A3B-MXFP4" in config_filename.name):
+    if current_platform.is_rocm() and (
+        "Qwen3.5-35B-A3B-MXFP4-AITER-TP2" in config_filename.name
+    ):
         from vllm.platforms.rocm import on_gfx950
 
-        if not on_gfx950() and "AITER-TP2" in config_filename.name:
+        if not on_gfx950():
             pytest.skip(
                 "Skipping Qwen3.5-35B-A3B-MXFP4-AITER-TP2 on non-GFX950 platforms. "
                 "The quantization scheme is not supported on non-GFX950 platforms."
             )
-        if not QUARK_MXFP4_TORCH_COMPATIBLE:
-            pytest.skip(
-                "Skipping Qwen3.5-35B-A3B-MXFP4: amd-quark >= 0.12 is required "
-                "on torch >= 2.11."
-            )
     # Parse server arguments from config (use shlex to handle quoted strings)
     server_args_str = eval_config.get("server_args", "")
     server_args = shlex.split(server_args_str) if server_args_str else []

tests/kernels/moe/test_ocp_mx_moe.py

@@ -1,30 +1,17 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
-import importlib.metadata
 import types
 from dataclasses import dataclass
-from importlib.util import find_spec
 
 import pytest
 import torch
-from packaging import version
 
 from tests.kernels.moe.utils import check_accuracy
 from vllm._aiter_ops import is_aiter_found, rocm_aiter_ops
 from vllm.platforms import current_platform
 from vllm.utils.flashinfer import has_flashinfer
 
-# MXFP4 via quark requires amd-quark >= 0.12 on torch >= 2.11.
-# Earlier torch releases work with older quark versions. See
-# https://github.com/amd/Quark/issues/34
-# TODO: Remove once amd-quark>=0.12.0
-QUARK_MXFP4_TORCH_COMPATIBLE = find_spec("quark") is not None and (
-    version.parse(importlib.metadata.version("amd-quark")) >= version.parse("0.12.0")
-    if version.parse(torch.__version__.split("+")[0]) >= version.parse("2.11")
-    else True
-)
-
 TRTLLM_GEN_MXFP4_AVAILABLE = (
     current_platform.is_cuda() and current_platform.is_device_capability_family(100)
 )
@@ -96,8 +83,8 @@ def enable_pickle(monkeypatch):
     ],
 )
 @pytest.mark.skipif(
-    not QUARK_MXFP4_TORCH_COMPATIBLE,
-    reason="MXFP4 via quark requires amd-quark >= 0.12 on torch >= 2.11.",
+    not ROCM_AVAILABLE,
+    reason="This test is only enabled on platforms that support amd-quark",
 )
 def test_mxfp4_loading_and_execution_moe(vllm_runner, model_case: ModelCase):
     if torch.accelerator.device_count() < model_case.tp:

tests/lora/test_gptoss_tp.py

@@ -1,12 +1,8 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
-import importlib.metadata
-from importlib.util import find_spec
 
 import pytest
-import torch
-from packaging import version
 
 import vllm
 from vllm.lora.request import LoRARequest
@@ -14,22 +10,6 @@ from vllm.platforms import current_platform
 
 from ..utils import multi_gpu_test
 
-# Require amd-quark >= 0.12 on torch >= 2.11.
-# Earlier torch releases work with older quark versions. See
-# https://github.com/amd/Quark/issues/34
-# TODO: Remove once amd-quark>=0.12.0
-QUARK_TORCH_COMPATIBLE = find_spec("quark") is not None and (
-    version.parse(importlib.metadata.version("amd-quark")) >= version.parse("0.12.0")
-    if version.parse(torch.__version__.split("+")[0]) >= version.parse("2.11")
-    else True
-)
-
-if current_platform.is_rocm() and not QUARK_TORCH_COMPATIBLE:
-    pytest.skip(
-        "This test requires amd-quark >= 0.12 on torch >= 2.11.",
-        allow_module_level=True,
-    )
-
 MODEL_PATH = "openai/gpt-oss-20b"
 
 PROMPT_TEMPLATE = """<|start|>system<|message|>You are ChatGPT, a large language model trained by OpenAI.
@@ -88,7 +68,9 @@ def generate_and_test(llm: vllm.LLM, lora_path: str, lora_id: int) -> None:
         generated_texts.append(generated_text)
         print(f"Prompt: {prompt!r}, Generated text: {generated_text!r}")
     for i in range(len(EXPECTED_LORA_OUTPUT)):
-        assert generated_texts[i].startswith(EXPECTED_LORA_OUTPUT[i])
+        generated = " ".join(generated_texts[i].split())
+        expected = " ".join(EXPECTED_LORA_OUTPUT[i].split())
+        assert generated.startswith(expected)
 
 
 # TODO: make the Mxfp4MoeBackend.TRITON spawn-safe.

tests/quantization/test_quark.py

@@ -41,7 +41,7 @@ else:
 from .reference_mxfp4 import dq_mxfp4_torch, qdq_mxfp4_torch
 
 # Minimum amd-quark version for MXFP4/OCP_MX tests (single source of truth).
-QUARK_MXFP4_MIN_VERSION = "0.8.99"
+QUARK_MXFP4_MIN_VERSION = "0.12"
 
 QUARK_MXFP4_AVAILABLE = find_spec("quark") is not None and version.parse(
     importlib.metadata.version("amd-quark")