PTQ-Level Efficiency, QAT-Level Accuracy | 训练后量化的效率,量化感知训练的精度
Update unit test script (#1290)
Sun, Xuehao8e6f337040ba
diff --git a/.azure-pipelines/docker/Dockerfile.devel b/.azure-pipelines/docker/Dockerfile.develindex 7f196142a..cbb4eda88 100644--- a/.azure-pipelines/docker/Dockerfile.devel+++ b/.azure-pipelines/docker/Dockerfile.devel@@ -43,7 +43,6 @@ USER hostuserENV PATH="/home/hostuser/.venv/bin:$PATH"ENV VIRTUAL_ENV="/home/hostuser/.venv"ENV UV_NO_PROGRESS=1 \- UV_COMPILE_BYTECODE=1 \UV_LINK_MODE=copyRUN uv venv --python=3.12 /home/hostuser/.venvdiff --git a/.azure-pipelines/scripts/ut/run_ut.sh b/.azure-pipelines/scripts/ut/run_ut.shindex 0e70c3794..73ef823b6 100644--- a/.azure-pipelines/scripts/ut/run_ut.sh+++ b/.azure-pipelines/scripts/ut/run_ut.sh@@ -13,12 +13,14 @@ uv pip install -r /auto-round/test/test_cpu/requirements.txt \# install latest gguf for ut testcd ~ || exit 1git clone -b master --quiet --single-branch https://github.com/ggml-org/llama.cpp.git && cd llama.cpp/gguf-py && uv pip install . sentencepiece+uv pip install --upgrade transformerscd /auto-round && uv pip install .echo "##[endgroup]"uv pip list+rm -rf /auto-round/auto_roundcd /auto-round/test || exit 1export LD_LIBRARY_PATH=${HOME}/.venv/lib/:$LD_LIBRARY_PATHdiff --git a/.azure-pipelines/scripts/ut/run_ut_cuda.sh b/.azure-pipelines/scripts/ut/run_ut_cuda.shindex ff796b94e..5ef83f527 100644--- a/.azure-pipelines/scripts/ut/run_ut_cuda.sh+++ b/.azure-pipelines/scripts/ut/run_ut_cuda.sh@@ -27,14 +27,14 @@ function create_conda_env() {# install AutoRoundcd ${REPO_PATH}- uv pip install torch==2.8.0 torchvision+ uv pip install torch==2.9.1 torchvisionuv pip install -r requirements.txtif [ -d "/proc/driver/nvidia" ]; thenexport PATH=/usr/local/cuda/bin${PATH:+:${PATH}}export LD_LIBRARY_PATH=$(python -c "import site; print(site.getsitepackages()[0])")/nvidia/nvjitlink/lib:$LD_LIBRARY_PATHfiuv pip install --no-build-isolation .- uv pip install pytest-cov pytest-html cmake==4.0.2+ uv pip install pytest-cov pytest-html cmake}function print_test_results_table() {@@ -42,9 +42,9 @@ function print_test_results_table() {local test_type=$2echo ""- echo "==========================================" >> ${SUMMARY_LOG}+ { printf '=%.0s' {1..120}; echo; } >> ${SUMMARY_LOG}echo "Test Results Summary - ${test_type}" >> ${SUMMARY_LOG}- echo "==========================================" >> ${SUMMARY_LOG}+ { printf '=%.0s' {1..120}; echo; } >> ${SUMMARY_LOG}printf "%-30s %-10s %-50s\n" "Test Case" "Result" "Log File" >> ${SUMMARY_LOG}printf "%-30s %-10s %-50s\n" "----------" "------" "--------" >> ${SUMMARY_LOG}@@ -81,9 +81,9 @@ function print_test_results_table() {fidone- echo "==========================================" >> ${SUMMARY_LOG}+ { printf '=%.0s' {1..120}; echo; } >> ${SUMMARY_LOG}printf "Total: %d, Passed: %d, Failed: %d\n" ${total_tests} ${passed_tests} ${failed_tests} >> ${SUMMARY_LOG}- echo "==========================================" >> ${SUMMARY_LOG}+ { printf '=%.0s' {1..120}; echo; } >> ${SUMMARY_LOG}echo "" >> ${SUMMARY_LOG}}@@ -95,12 +95,13 @@ function run_unit_test() {rm -rf .coverage* *.xml *.htmluv pip install -v git+https://github.com/casper-hansen/AutoAWQ.git --no-build-isolation- uv pip install -v git+https://github.com/ModelCloud/[email protected] --no-build-isolation+ uv pip install -v git+https://github.com/ModelCloud/[email protected] --no-build-isolationuv pip install -r https://raw.githubusercontent.com/ModelCloud/GPTQModel/refs/heads/main/requirements.txtCMAKE_ARGS="-DGGML_CUDA=on -DLLAVA_BUILD=off" uv pip install llama-cpp-pythonuv pip install 'git+https://github.com/ggml-org/llama.cpp.git#subdirectory=gguf-py'uv pip install -r test_cuda/requirements.txtuv pip install -r test_cuda/requirements_diffusion.txt+ uv pip install -r test_cuda/requirements_sglang.txtpip list > ${LOG_DIR}/ut_pip_list.txtexport COVERAGE_RCFILE=${REPO_PATH}/.azure-pipelines/scripts/ut/.coveragediff --git a/.azure-pipelines/scripts/ut/run_ut_hpu.sh b/.azure-pipelines/scripts/ut/run_ut_hpu.shindex b71a19c43..f90fdb7c0 100644--- a/.azure-pipelines/scripts/ut/run_ut_hpu.sh+++ b/.azure-pipelines/scripts/ut/run_ut_hpu.sh@@ -7,6 +7,7 @@ export TQDM_MININTERVAL=60pip install pytest-cov pytest-htmlpip list+rm -rf /auto-round/auto_roundcd /auto-round/test || exit 1export LD_LIBRARY_PATH=/usr/local/lib/:$LD_LIBRARY_PATHdiff --git a/.azure-pipelines/scripts/ut/run_ut_xpu.sh b/.azure-pipelines/scripts/ut/run_ut_xpu.shindex 7716a0cc6..b1549d73e 100644--- a/.azure-pipelines/scripts/ut/run_ut_xpu.sh+++ b/.azure-pipelines/scripts/ut/run_ut_xpu.sh@@ -11,6 +11,7 @@ echo "##[endgroup]"uv pip list# test ark cpu part only before external xpu available+rm -rf /auto-round/auto_roundcd /auto-round/test || exit 1export LD_LIBRARY_PATH=${HOME}/.venv/lib/:$LD_LIBRARY_PATHdiff --git a/test/test_cuda/requirements.txt b/test/test_cuda/requirements.txtindex 071eb233e..fa36ffcc4 100644--- a/test/test_cuda/requirements.txt+++ b/test/test_cuda/requirements.txt@@ -1,12 +1,11 @@# pip install -v git+https://github.com/casper-hansen/AutoAWQ.git --no-build-isolationauto-gptqeinops-# pip install -v git+https://github.com/ModelCloud/[email protected] --no-build-isolation-intel-extension-for-pytorch+# pip install -v git+https://github.com/ModelCloud/GPTQModel.git --no-build-isolationlm-eval>=0.4.9.1optimumpandaspillowtorchvisionnumba-vllm>=0.8.5.post1,<=0.11.0+vllm>=0.8.5.post1