Skip to content

PTQ-Level Efficiency, QAT-Level Accuracy | 训练后量化的效率,量化感知训练的精度

Add script for cuda unit test (#567)

Sun, Xuehao4e617551f8d8
diff --git a/.azure-pipelines/scripts/ut/run_ut_cuda.sh b/.azure-pipelines/scripts/ut/run_ut_cuda.sh
new file mode 100644
index 000000000..02073e958
--- /dev/null
+++ b/.azure-pipelines/scripts/ut/run_ut_cuda.sh
@@ -0,0 +1,112 @@
+#!/bin/bash
+set -xe
+
+CONDA_ENV_NAME="unittest_cuda"
+PYTHON_VERSION="3.10"
+REPO_PATH=$(git rev-parse --show-toplevel)
+LOG_DIR=${REPO_PATH}/ut_log_dir
+[[ -z "$CUDA_VISIBLE_DEVICES" ]] && export CUDA_VISIBLE_DEVICES=0
+
+function create_conda_env() {
+ echo "-----[VAL INFO] create conda env -----"
+ [[ -d ${HOME}/anaconda3/bin ]] && export PATH=${HOME}/anaconda3/bin/:$PATH
+ [[ -d ${HOME}/miniforge3/bin ]] && export PATH=${HOME}/miniforge3/bin/:$PATH
+ [[ -d ${HOME}/miniconda3/bin ]] && export PATH=${HOME}/miniconda3/bin/:$PATH
+
+ # create conda env
+ source activate base
+ if conda info --envs | grep -q "^$CONDA_ENV_NAME\s"; then conda remove -n ${CONDA_ENV_NAME} --all -y; fi
+ conda create -n ${CONDA_ENV_NAME} python=${PYTHON_VERSION} setuptools=69.5.1 -y
+ source activate ${CONDA_ENV_NAME}
+ conda install -c conda-forge git gxx=11.2.0 gcc=11.2.0 gdb sysroot_linux-64 libgcc uv -y
+ export LD_PRELOAD=${CONDA_PREFIX}/lib/libstdc++.so.6
+
+ # install AutoRound
+ cd ${REPO_PATH}
+ pip uninstall auto-round -y
+ uv pip install -r requirements.txt
+ sed -i '/^torch==/d;/^transformers==/d;/^lm-eval==/d' requirements.txt
+ if [ -d "/proc/driver/nvidia" ]; then
+ export PATH=/usr/local/cuda/bin${PATH:+:${PATH}}
+ export LD_LIBRARY_PATH=$(python -c "import site; print(site.getsitepackages()[0])")/nvidia/nvjitlink/lib:$LD_LIBRARY_PATH
+ fi
+ uv pip install -v --no-build-isolation .
+ uv pip install pytest-cov pytest-html cmake==4.0.2
+}
+
+function run_unit_test() {
+ # install unit test dependencies
+ create_conda_env
+
+ cd ${REPO_PATH}/test/test_cuda
+ rm -rf .coverage* *.xml *.html
+
+ uv pip install -v git+https://github.com/casper-hansen/AutoAWQ.git --no-build-isolation
+ uv pip install -v git+https://github.com/ModelCloud/[email protected] --no-build-isolation
+ CMAKE_ARGS="-DGGML_CUDA=on -DLLAVA_BUILD=off" uv pip install llama-cpp-python
+ uv pip install 'git+https://github.com/ggml-org/llama.cpp.git#subdirectory=gguf-py'
+ uv pip install -r requirements.txt
+
+ uv pip list
+ export COVERAGE_RCFILE=${REPO_PATH}/.azure-pipelines/scripts/ut/.coverage
+ local auto_round_path=$(python -c 'import auto_round; print(auto_round.__path__[0])')
+
+ # setup test env
+ mkdir -p ${LOG_DIR}
+ local ut_log_name=${LOG_DIR}/unittest_cuda.log
+ find . -name "test_*.py" | sed "s,\.\/,python -m pytest --cov=\"${auto_round_path}\" --cov-report term --html=report.html --self-contained-html --cov-report xml:coverage.xml --cov-append -vs --disable-warnings ,g" >run.sh
+ cat run.sh
+
+ # run unit test
+ bash run.sh 2>&1 | tee ${ut_log_name}
+
+ cp report.html ${LOG_DIR}/
+ cp coverage.xml ${LOG_DIR}/
+
+ if [ $(grep -c '== FAILURES ==' ${ut_log_name}) != 0 ] || [ $(grep -c '== ERRORS ==' ${ut_log_name}) != 0 ] || [ $(grep -c ' passed' ${ut_log_name}) == 0 ]; then
+ echo "Find errors in pytest case, please check the output..."
+ fi
+}
+
+function run_unit_test_vlm() {
+ # install unit test dependencies
+ create_conda_env
+ cd ${REPO_PATH}/test/test_cuda
+ rm -rf .coverage* *.xml *.html
+
+ uv pip install git+https://github.com/haotian-liu/[email protected] --no-deps
+ local site_path=$(python -c "import site; print(site.getsitepackages()[0])")
+ # reference https://github.com/haotian-liu/LLaVA/issues/1448#issuecomment-2119845242
+ sed -i '/inputs\[.*image_sizes.*\] = image_sizes/a\ inputs.pop("cache_position")' ${site_path}/llava/model/language_model/llava_llama.py
+ uv pip install git+https://github.com/deepseek-ai/DeepSeek-VL2.git timm attrdict --no-deps
+ uv pip install -v git+https://github.com/casper-hansen/[email protected] --no-build-isolation
+ uv pip install flash-attn==2.7.4.post1 --no-build-isolation
+ uv pip install -r requirements_vlm.txt
+
+ uv pip list
+ export COVERAGE_RCFILE=${REPO_PATH}/.azure-pipelines/scripts/ut/.coverage
+ local auto_round_path=$(python -c 'import auto_round; print(auto_round.__path__[0])')
+
+ # setup test env
+ mkdir -p ${LOG_DIR}
+ local ut_log_name=${LOG_DIR}/unittest_cuda_vlm.log
+ find . -name "test*vlms.py" | sed "s,\.\/,python -m pytest --cov=\"${auto_round_path}\" --cov-report term --html=report_vlms.html --self-contained-html --cov-report xml:coverage_vlms.xml --cov-append -vs --disable-warnings ,g" >run_vlms.sh
+ cat run_vlms.sh
+
+ # run unit test
+ bash run_vlms.sh 2>&1 | tee ${ut_log_name}
+
+ cp report_vlms.html ${LOG_DIR}/
+ cp coverage_vlms.xml ${LOG_DIR}/
+
+ if [ $(grep -c '== FAILURES ==' ${ut_log_name}) != 0 ] || [ $(grep -c '== ERRORS ==' ${ut_log_name}) != 0 ] || [ $(grep -c ' passed' ${ut_log_name}) == 0 ]; then
+ echo "Find errors in pytest case, please check the output..."
+ fi
+}
+
+function main() {
+ run_unit_test
+ run_unit_test_vlm
+}
+
+main
diff --git a/.azure-pipelines/scripts/ut/run_ut_hpu.sh b/.azure-pipelines/scripts/ut/run_ut_hpu.sh
index ec2ad0d20..b2f895c09 100644
--- a/.azure-pipelines/scripts/ut/run_ut_hpu.sh
+++ b/.azure-pipelines/scripts/ut/run_ut_hpu.sh
@@ -3,6 +3,7 @@ set -xe
# install requirements
echo "set up UT env..."
+export TQDM_MININTERVAL=60
pip install pytest-cov pytest-html
pip list
diff --git a/.gitignore b/.gitignore
index 4226ce31e..bfb194c6d 100644
--- a/.gitignore
+++ b/.gitignore
@@ -9,3 +9,4 @@ dist/
.clangd
CMakeUserPresets.json
tmp_autoround/
+ut_log_dir/
diff --git a/test/test_cuda/requirements_vlm.txt b/test/test_cuda/requirements_vlm.txt
index 0bfb0c39f..b1b06b14c 100644
--- a/test/test_cuda/requirements_vlm.txt
+++ b/test/test_cuda/requirements_vlm.txt
@@ -1,8 +1,8 @@
-# git+https://github.com/haotian-liu/[email protected]
-# pip install git+https://github.com/deepseek-ai/DeepSeek-VL2.git
+# git+https://github.com/haotian-liu/[email protected] --no-deps
+# pip install git+https://github.com/deepseek-ai/DeepSeek-VL2.git --no-deps
# pip install -v git+https://github.com/casper-hansen/[email protected] --no-build-isolation
+# pip install flash-attn==2.7.4.post1 --no-build-isolation
accelerate
-# autoawq
bitsandbytes
datasets
einops
diff --git a/test/test_cuda/test_support_vlms.py b/test/test_cuda/test_support_vlms.py
index 4fac40c2c..36c5f2082 100644
--- a/test/test_cuda/test_support_vlms.py
+++ b/test/test_cuda/test_support_vlms.py
@@ -24,7 +24,7 @@ class TestSupportVLMS(unittest.TestCase):
def tearDownClass(self):
shutil.rmtree(self.save_dir, ignore_errors=True)
- @require_gptqmodel
+ # @require_gptqmodel
# def test_qwen2(self):
# model_path = "/models/Qwen2-VL-2B-Instruct/"
# # test tune
@@ -217,7 +217,7 @@ class TestSupportVLMS(unittest.TestCase):
print(tokenizer.batch_decode(output))
shutil.rmtree(quantized_model_path, ignore_errors=True)
- @require_gptqmodel
+ # @require_gptqmodel
# def test_llama(self):
# model_path = "/models/Llama-3.2-11B-Vision-Instruct/"
# ## test tune