Update torch to 2.11.0 in CI

.azure-pipelines/scripts/cuda_unit_test/run_cuda_ut.sh

@@ -62,22 +62,9 @@ function run_unit_test() {
     # install unit test dependencies
     echo "##[group]set up UT env..."
     cd "${BUILD_SOURCESDIRECTORY}" || exit 1
-    uv pip install torch==2.10.0 torchvision
-    uv pip install git+https://github.com/casper-hansen/AutoAWQ.git --no-build-isolation
-
-    # install gptqmodel
-    CUDA_VER=$(python -c 'import torch; print(f"cu{torch.version.cuda.replace(".", "")}")')
-    PY_VER=$(python -c 'import sys; print(f"cp{sys.version_info.major}{sys.version_info.minor}")')
-    TORCH_VER="torch2.10"
-    WHEEL="gptqmodel-5.7.0-${CUDA_VER}${TORCH_VER}-${PY_VER}-${PY_VER}-linux_x86_64.whl"
-    URL="https://pkgs.dev.azure.com/lpot-inc/b7121868-d73a-4794-90c1-23135f974d09/_packaging/4728fbab-e069-4cbd-bcca-d35f4d42256b/pypi/download/gptqmodel/5.7/${WHEEL}"
-    wget -q "$URL" -O "$WHEEL" || { echo "Download failed. Check CUDA/PyTorch/Python versions match (cu126/cu128/cu130, torch2.10, cp310-cp313)"; exit 1; }
-    mv "$WHEEL" "${WHEEL/-${CUDA_VER}${TORCH_VER}-/+${CUDA_VER}.${TORCH_VER}-}"
-    uv pip install "./${WHEEL/-${CUDA_VER}${TORCH_VER}-/+${CUDA_VER}.${TORCH_VER}-}" --no-build-isolation
-    rm -f "./${WHEEL/-${CUDA_VER}${TORCH_VER}-/+${CUDA_VER}.${TORCH_VER}-}"
-
-    uv pip install gptqmodel --extra-index-url https://pkgs.dev.azure.com/lpot-inc/neural-compressor/_packaging/gptqmodel-wheels/pypi/simple/
-    uv pip install -r https://raw.githubusercontent.com/ModelCloud/GPTQModel/refs/tags/v5.7.0/requirements.txt
+    uv pip install torch==2.11.0 torchvision --index-url https://download.pytorch.org/whl/cu128
+    uv pip install https://github.com/XuehaoSun/GPTQModel/releases/download/v5.8.0/gptqmodel-5.8.0+cu128torch2.11-cp312-cp312-linux_x86_64.whl
+    uv pip install -r https://raw.githubusercontent.com/ModelCloud/GPTQModel/refs/tags/v5.8.0/requirements.txt
     uv pip install https://github.com/XuehaoSun/llama-cpp-python/releases/download/v0.3.16/llama_cpp_python-0.3.16-cp312-cp312-linux_x86_64.whl
     uv pip install 'git+https://github.com/ggml-org/llama.cpp.git#subdirectory=gguf-py'
     uv pip install -r test/test_cuda/requirements.txt
@@ -126,8 +113,9 @@ function run_unit_test_llmc() {
     uv pip install -U pytest-cov pytest-html
     uv pip install -r test/test_cuda/requirements_llmc.txt
     uv pip install .
-    echo "##[endgroup]"
     uv pip list
+    echo "##[endgroup]"
+
     cd "${BUILD_SOURCESDIRECTORY}/test" || exit 1
 
     export COVERAGE_RCFILE="${BUILD_SOURCESDIRECTORY}/.azure-pipelines/scripts/ut/.coverage"
@@ -151,9 +139,9 @@ function run_unit_test_sglang() {
     uv pip install -U pytest-cov pytest-html
     uv pip install -r test/test_cuda/requirements_sglang.txt
     uv pip install .
+    uv pip list
     echo "##[endgroup]"
 
-    uv pip list
     cd "${BUILD_SOURCESDIRECTORY}/test" || exit 1
     export COVERAGE_RCFILE="${BUILD_SOURCESDIRECTORY}/.azure-pipelines/scripts/ut/.coverage"
 
@@ -176,9 +164,9 @@ function run_unit_test_vllm() {
     uv pip install -U pytest-cov pytest-html
     uv pip install -r test/test_cuda/requirements_vllm.txt
     uv pip install .
+    uv pip list
     echo "##[endgroup]"
 
-    uv pip list
     cd "${BUILD_SOURCESDIRECTORY}/test" || exit 1
     export COVERAGE_RCFILE="${BUILD_SOURCESDIRECTORY}/.azure-pipelines/scripts/ut/.coverage"
 
@@ -197,16 +185,14 @@ function main() {
     setup_environment
     if [ "${test_case}" == "vlm" ]; then
         run_unit_test_vlm
-    elif [ "${test_case}" == "llmc" ]; then
-        run_unit_test_llmc
-    elif [ "${test_case}" == "sglang" ]; then
+    elif [ "${test_case}" == "specific" ]; then
         run_unit_test_sglang
-    elif [ "${test_case}" == "vllm" ]; then
+        run_unit_test_llmc
         run_unit_test_vllm
     elif [ "${test_case}" == "all" ]; then
         run_unit_test
     else
-        echo "##[error]Invalid test case specified: ${test_case}. Please use 'vlm', 'llmc', 'sglang', 'vllm', or 'all'."
+        echo "##[error]Invalid test case specified: ${test_case}. Please use 'vlm', 'specific', or 'all'."
         exit 1
     fi
     check_storage_usage

.azure-pipelines/scripts/cuda_unit_test/runpod_manager.py

@@ -94,15 +94,27 @@ def check_gpu_count(token):
 def run_create_pod(api_key, payload):
     url = "https://rest.runpod.io/v1/pods"
     headers = {"Authorization": f"Bearer {api_key}", "Content-Type": "application/json"}
-    response = requests.post(url, json=payload, headers=headers)
+    max_retries = 3
+
+    for attempt in range(max_retries + 1):
+        response = requests.post(url, json=payload, headers=headers)
+
+        if response.status_code >= 500:
+            if attempt < max_retries:
+                print(f"⚠️ {response.status_code} Error, Retrying in 60 seconds ({attempt + 1}/{max_retries})...")
+                time.sleep(60)
+                continue
+            else:
+                print(f"❌ {response.status_code} Error, Reached maximum retry attempts ({max_retries}), giving up.")
+        response.raise_for_status()
 
-    response.raise_for_status()
-    result = response.json()
-    if "errors" in result:
-        print("❌ Errors:")
-        print(json.dumps(result["errors"], indent=2))
-        sys.exit(1)
-    return result
+        result = response.json()
+        if "errors" in result:
+            print("❌ Errors:")
+            print(json.dumps(result["errors"], indent=2))
+            sys.exit(1)
+
+        return result
 
 
 def create_pod(args):

.azure-pipelines/scripts/ut/run_ut_cuda.sh

@@ -91,8 +91,7 @@ function run_unit_test() {
     cd ${REPO_PATH}/test
     rm -rf .coverage* *.xml *.html
 
-    uv pip install torch==2.10.0 torchvision
-    uv pip install -v git+https://github.com/casper-hansen/AutoAWQ.git --no-build-isolation
+    uv pip install torch==2.11.0 torchvision --index-url https://download.pytorch.org/whl/cu128
     uv pip install gptqmodel --no-build-isolation
     uv pip install -r https://raw.githubusercontent.com/ModelCloud/GPTQModel/refs/heads/main/requirements.txt
     CMAKE_ARGS="-DGGML_CUDA=on -DLLAVA_BUILD=off" uv pip install llama-cpp-python
@@ -132,14 +131,11 @@ function run_unit_test_vlm() {
     cd ${REPO_PATH}/test
     rm -rf .coverage* *.xml *.html
 
-    uv pip install torch==2.10.0 torchvision
+    uv pip install torch==2.11.0 torchvision --index-url https://download.pytorch.org/whl/cu128
+    uv pip install gptqmodel --no-build-isolation
     uv pip install git+https://github.com/haotian-liu/LLaVA.git@v1.2.2 --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/AutoAWQ.git@v0.2.0 --no-build-isolation
-    uv pip install flash-attn==2.7.4.post1 --no-build-isolation
+    uv pip install flash-attn==2.8.3 --no-build-isolation
     uv pip install -r test_cuda/requirements_vlm.txt
     cd ${REPO_PATH} && uv pip install . && cd ${REPO_PATH}/test
 

.azure-pipelines/template/ut-template.yml

@@ -46,17 +46,17 @@ steps:
     - script: |
         if [ "${{ parameters.utScriptFileName }}" == "run_ut_xpu" ];then
           docker exec ${{ parameters.utContainerName }} bash -c "cd /auto-round \
-            && uv pip install torch==2.10.0 torchvision --index-url https://download.pytorch.org/whl/xpu \
-            && uv pip install torch==2.10.0 auto-round-lib \
+            && uv pip install torch==2.11.0 torchvision --index-url https://download.pytorch.org/whl/xpu \
+            && uv pip install torch==2.11.0 auto-round-lib \
             && uv pip install -r requirements.txt \
             && uv pip install -r test/test_ark/requirements.txt \
             && uv pip install -r test/test_xpu/requirements.txt \
             && cd /auto-round && uv pip install . \
             && uv pip list"
         else 
           docker exec ${{ parameters.utContainerName }} bash -c "cd /auto-round \
-            && uv pip install torch==2.10.0 torchvision --index-url https://download.pytorch.org/whl/cpu \
-            && uv pip install torch==2.10.0 auto-round-lib \
+            && uv pip install torch==2.11.0 torchvision --index-url https://download.pytorch.org/whl/cpu \
+            && uv pip install torch==2.11.0 auto-round-lib \
             && uv pip install -r requirements.txt \
             && uv pip install -r requirements-cpu.txt \
             && uv pip install -r test/test_cpu/requirements.txt \

.azure-pipelines/unit-test-cuda.yml

@@ -138,11 +138,7 @@ stages:
                   if [ ${{ pair.value.PART }} -eq 0 ]; then
                     export UV_NO_CACHE=0
                     bash .azure-pipelines/scripts/cuda_unit_test/run_cuda_ut.sh \
-                      --test_case=sglang
-                    bash .azure-pipelines/scripts/cuda_unit_test/run_cuda_ut.sh \
-                      --test_case=llmc
-                    bash .azure-pipelines/scripts/cuda_unit_test/run_cuda_ut.sh \
-                      --test_case=vllm
+                      --test_case=specific
                   else
                     bash .azure-pipelines/scripts/cuda_unit_test/run_cuda_ut.sh \
                       --test_case=all \