Update torch to 2.11.0 in CI

[XuehaoSun]

Description

Please briefly describe your main changes, the motivation.

Type of Change

• Bug fix
• New feature
• Documentation update
• Performance improvement
• Code refactoring
• Other (please specify):

Related Issues

Fixes or relates to #

Checklist Before Submitting

• My code has been tested locally.
• Documentation has been updated as needed.
• New or updated tests are included where applicable.

[Copilot]

Pull request overview

Updates the Azure Pipelines CI environment to use newer PyTorch tooling (Torch 2.11.0) and adjusts CUDA CI orchestration to align with the updated dependency setup.

Changes:

• Bump Torch from 2.10.0 to 2.11.0 in multiple CI/UT setup paths (CPU/XPU/CUDA).
• Consolidate CUDA “specific” test execution into a single invocation for part0.
• Add basic retry logic for RunPod pod creation; update CUDA UT dependency installs (e.g., gptqmodel wheel, flash-attn version).

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
.azure-pipelines/unit-test-cuda.yml	Switch part0 CUDA run to a single --test_case=specific invocation.
.azure-pipelines/template/ut-template.yml	Update Torch pin to 2.11.0 for CPU/XPU container UT environment setup.
.azure-pipelines/scripts/ut/run_ut_cuda.sh	Update CUDA UT dependency installation (Torch/cu128 index URL; VLM deps/version adjustments).
.azure-pipelines/scripts/cuda_unit_test/runpod_manager.py	Add retry loop around RunPod pod creation request.
.azure-pipelines/scripts/cuda_unit_test/run_cuda_ut.sh	Update Torch/gptqmodel/llama-cpp-python installs and introduce specific test mode.

Comments suppressed due to low confidence (1)

.azure-pipelines/scripts/ut/run_ut_cuda.sh:98

• The CUDA UT conda environment no longer installs an AWQ provider (AutoAWQ), but there are CUDA tests decorated with require_awq (expects the 'awq' module). As a result, those tests will be skipped when this script is used, reducing effective test coverage for AWQ-related code paths. If this script is intended to validate AWQ functionality, re-add an AWQ provider install step or update the test plan/markers accordingly.

    cd ${REPO_PATH}/test
    rm -rf .coverage* *.xml *.html

    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
    uv pip install 'git+https://github.com/ggml-org/llama.cpp.git#subdirectory=gguf-py'

[XuehaoSun]

/azp run Unit-Test-CUDA-AutoRound

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).