PTQ-Level Efficiency, QAT-Level Accuracy | 训练后量化的效率,量化感知训练的精度
Fix vllm CUDA CI (#1750)
Sun, Xuehao5e2dbdbe3ced
diff --git a/.azure-pipelines/scripts/cuda_unit_test/run_cuda_ut.sh b/.azure-pipelines/scripts/cuda_unit_test/run_cuda_ut.shindex 68948e0d3..415635f52 100644--- a/.azure-pipelines/scripts/cuda_unit_test/run_cuda_ut.sh+++ b/.azure-pipelines/scripts/cuda_unit_test/run_cuda_ut.sh@@ -163,7 +163,10 @@ function run_unit_test_vllm() {rm -rf /root/.venvuv venv --python=3.12 /root/.venvuv pip install -U pytest-cov pytest-html- uv pip install -r test/test_cuda/requirements_vllm.txt --extra-index-url https://download.pytorch.org/whl/cu128 --index-strategy unsafe-best-match+ uv pip install -r test/test_cuda/requirements_vllm.txt \+ --extra-index-url https://wheels.vllm.ai/0.20.0/cu129 \+ --extra-index-url https://download.pytorch.org/whl/cu128 \+ --index-strategy unsafe-best-matchuv pip install .uv pip listecho "##[endgroup]"diff --git a/.azure-pipelines/scripts/ut/run_ut_cuda.sh b/.azure-pipelines/scripts/ut/run_ut_cuda.shindex 770da49c5..ce24916e3 100644--- a/.azure-pipelines/scripts/ut/run_ut_cuda.sh+++ b/.azure-pipelines/scripts/ut/run_ut_cuda.sh@@ -236,7 +236,10 @@ function run_unit_test_vllm() {cd ${REPO_PATH}/testrm -rf .coverage* *.xml *.html- uv pip install -r test_cuda/requirements_vllm.txt --extra-index-url https://download.pytorch.org/whl/cu128 --index-strategy unsafe-best-match+ uv pip install -r test_cuda/requirements_vllm.txt \+ --extra-index-url https://wheels.vllm.ai/0.20.0/cu129 \+ --extra-index-url https://download.pytorch.org/whl/cu128 \+ --index-strategy unsafe-best-matchcd ${REPO_PATH} && uv pip install . && cd ${REPO_PATH}/testpip list > ${LOG_DIR}/vllm_ut_pip_list.txt