[perf] Fun-ASR-Nano fine-tuning on B200: 0.244 s to 0.068 s per step at 1 GPU, mostly from cuDNN SDPA plan builds on every new batch shape #35
Workflow file for this run
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| name: Validate Fun-ASR-Nano vLLM | |
| on: | |
| pull_request: | |
| paths: | |
| - "funasr/models/fun_asr_nano/**" | |
| - "funasr/models/glm_asr/**" | |
| - "tests/test_fun_asr_nano_*.py" | |
| - "tests/test_glm_asr_vllm_*.py" | |
| - "tests/test_vllm_model_source_docs.py" | |
| - ".github/workflows/test-fun-asr-nano-vllm.yml" | |
| push: | |
| branches: | |
| - main | |
| paths: | |
| - "funasr/models/fun_asr_nano/**" | |
| - "funasr/models/glm_asr/**" | |
| - "tests/test_fun_asr_nano_*.py" | |
| - "tests/test_glm_asr_vllm_*.py" | |
| - "tests/test_vllm_model_source_docs.py" | |
| - ".github/workflows/test-fun-asr-nano-vllm.yml" | |
| permissions: | |
| contents: read | |
| jobs: | |
| contract: | |
| runs-on: ubuntu-latest | |
| timeout-minutes: 20 | |
| steps: | |
| - uses: actions/checkout@v4 | |
| - uses: actions/setup-python@v5 | |
| with: | |
| python-version: "3.11" | |
| cache: pip | |
| - name: Install contract-test dependencies | |
| run: | | |
| python -m pip install --upgrade pip | |
| python -m pip install \ | |
| torch --index-url https://download.pytorch.org/whl/cpu | |
| python -m pip install \ | |
| pytest numpy soundfile requests packaging \ | |
| omegaconf hydra-core kaldiio librosa tqdm pyyaml | |
| - name: Check syntax | |
| run: | | |
| python -m py_compile \ | |
| funasr/models/fun_asr_nano/inference_vllm.py \ | |
| funasr/models/fun_asr_nano/inference_vllm_streaming.py | |
| - name: Run vLLM regression contracts | |
| run: | | |
| python -m pytest -q \ | |
| tests/test_fun_asr_nano_autocast_device.py \ | |
| tests/test_fun_asr_nano_ctc_batch_fallback.py \ | |
| tests/test_fun_asr_nano_lora_injection.py \ | |
| tests/test_fun_asr_nano_missing_ctc_weights.py \ | |
| tests/test_fun_asr_nano_openai_response.py \ | |
| tests/test_fun_asr_nano_repetition_penalty.py \ | |
| tests/test_fun_asr_nano_vllm_dtype.py \ | |
| tests/test_glm_asr_vllm_dedup_keys.py \ | |
| tests/test_glm_asr_vllm_dtype.py \ | |
| tests/test_glm_asr_vllm_sampling.py \ | |
| tests/test_vllm_model_source_docs.py |