是否支持华为Ascend npu
Notice: In order to resolve issues more efficiently, please raise issue following the template. (注意:为了更加高效率解决您遇到的问题,请按照模板提问,补充细节)
❓ Questions and Help
请问是否有支持华为Ascend npu的计划?
Before asking:
- search the issues.
- search the docs.
What is your question?
Code
What have you tried?
What's your environment?
- OS (e.g., Linux):
- FunASR Version (e.g., 1.0.0):
- ModelScope Version (e.g., 1.11.0):
- PyTorch Version (e.g., 2.0.0):
- How you installed funasr (
pip, source): - Python version:
- GPU (e.g., V100M32)
- CUDA/cuDNN version (e.g., cuda11.7):
- Docker version (e.g., funasr-runtime-sdk-cpu-0.4.1)
- Any other relevant information:
同问
同问
同问
各位可以参考下我 fork 后的代码修改和文档:
- https://github.com/yzztin/SenseVoice-npu
- https://github.com/yzztin/SenseVoice-npu/blob/npu-support/README_npu.md
各位可以参考下我 fork 后的代码修改和文档:
- https://github.com/yzztin/SenseVoice-npu
- https://github.com/yzztin/SenseVoice-npu/blob/npu-support/README_npu.md
不用torchAir的话,性能太差了
不用torchAir的话,性能太差了
这里有官方适配的文档,供参考:
- https://gitee.com/ascend/ModelZoo-PyTorch/blob/master/ACL_PyTorch/built-in/audio/SenseVoice/TorchAir/README.md
- https://gitee.com/ascend/ModelZoo-PyTorch/blob/master/ACL_PyTorch/built-in/audio/SenseVoice/README_onnx.md
对的,用官方torchAir的版本简单跑了一下性能测试,你的版本跟torchAir有几倍的性能差距
请问官方版本可以直接在910B上运行推理吗
可以关注 https://github.com/k2-fsa/sherpa-onnx/pull/2697
近期会支持 sense voice + Ascend 910B NPU.
paraformer 的支持更难,我们都搞定了. SenseVoice 更简单,所以肯定能支持。
https://github.com/k2-fsa/sherpa-onnx/pull/2707
刚把 sense voice 导出到 ascend npu 910B.
https://github.com/k2-fsa/sherpa-onnx/pull/2728
增加了 ACL C API 的例子。
已经更新文档
https://k2-fsa.github.io/sherpa/onnx/ascend/index.html