[Core] Deprecate xformers (#29262)

Signed-off-by: Roger Wang <hey@rogerw.io>
This commit is contained in:
Roger Wang
2025-11-24 04:18:55 +00:00
committed by GitHub
parent 5253f4276f
commit 0ff70821c9
31 changed files with 77 additions and 963 deletions
-12
View File
@@ -57,10 +57,6 @@ def do_sample(llm: vllm.LLM, lora_path: str, lora_id: int) -> list[str]:
return generated_texts
@pytest.mark.xfail(
current_platform.is_rocm(),
reason="MiniCPM-V dependency xformers incompatible with ROCm",
)
def test_minicpmv_lora(minicpmv_lora_files):
llm = vllm.LLM(
MODEL_PATH,
@@ -84,10 +80,6 @@ def test_minicpmv_lora(minicpmv_lora_files):
@pytest.mark.skipif(
current_platform.is_cuda_alike(), reason="Skipping to avoid redundant model tests"
)
@pytest.mark.xfail(
current_platform.is_rocm(),
reason="MiniCPM-V dependency xformers incompatible with ROCm",
)
@multi_gpu_test(num_gpus=4)
def test_minicpmv_tp4_wo_fully_sharded_loras(minicpmv_lora_files):
llm = vllm.LLM(
@@ -108,10 +100,6 @@ def test_minicpmv_tp4_wo_fully_sharded_loras(minicpmv_lora_files):
@pytest.mark.skipif(
current_platform.is_cuda_alike(), reason="Skipping to avoid redundant model tests"
)
@pytest.mark.xfail(
current_platform.is_rocm(),
reason="MiniCPM-V dependency xformers incompatible with ROCm",
)
@multi_gpu_test(num_gpus=4)
def test_minicpmv_tp4_fully_sharded_loras(minicpmv_lora_files):
llm = vllm.LLM(
-15
View File
@@ -2,12 +2,9 @@
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
from dataclasses import dataclass
import pytest
import vllm
from vllm.assets.image import ImageAsset
from vllm.lora.request import LoRARequest
from vllm.platforms import current_platform
from vllm.sampling_params import BeamSearchParams
@@ -142,10 +139,6 @@ QWEN2VL_MODEL_PATH = "Qwen/Qwen2-VL-2B-Instruct"
QWEN25VL_MODEL_PATH = "Qwen/Qwen2.5-VL-3B-Instruct"
@pytest.mark.xfail(
current_platform.is_rocm(),
reason="Qwen2-VL dependency xformers incompatible with ROCm",
)
def test_qwen2vl_lora(qwen2vl_lora_files):
"""Test Qwen 2.0 VL model with LoRA"""
config = TestConfig(model_path=QWEN2VL_MODEL_PATH, lora_path=qwen2vl_lora_files)
@@ -156,10 +149,6 @@ def test_qwen2vl_lora(qwen2vl_lora_files):
tester.run_test(TEST_IMAGES, expected_outputs=EXPECTED_OUTPUTS, lora_id=lora_id)
@pytest.mark.xfail(
current_platform.is_rocm(),
reason="Qwen2-VL dependency xformers incompatible with ROCm",
)
def test_qwen2vl_lora_beam_search(qwen2vl_lora_files):
"""Test Qwen 2.0 VL model with LoRA through beam search."""
config = TestConfig(model_path=QWEN2VL_MODEL_PATH, lora_path=qwen2vl_lora_files)
@@ -178,10 +167,6 @@ def test_qwen2vl_lora_beam_search(qwen2vl_lora_files):
)
@pytest.mark.xfail(
current_platform.is_rocm(),
reason="Qwen2.5-VL dependency xformers incompatible with ROCm",
)
def test_qwen25vl_lora(qwen25vl_lora_files):
"""Test Qwen 2.5 VL model with LoRA"""
config = TestConfig(model_path=QWEN25VL_MODEL_PATH, lora_path=qwen25vl_lora_files)