-
Notifications
You must be signed in to change notification settings - Fork 1.1k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[paraformer] When is ONNX GPU export supported. #2503
Comments
关注一下cif那部分的转onnx |
是的,确实主要是CIF那块儿的问题;
|
第一部分会重构下这个函数 这个函数也会影响torch.compile等函数 第二部分 onnx 是支持for 循环导出的这里我有时间整一下, 你可以搜一下torch for loop to onnx; 或者可以对齐下parallel cif的实现 提个pr过来 |
@whisper-yu #2515 帮忙mask试下这个 🙏 |
感谢!那我这里先试一下onnx for循环导出的问题,看能不能解决; 但是推理耗时不稳定的问题,会是CIF这块儿的问题吗? |
应该是 其他结构都类transformer 推理应该很稳定 |
我做了一些测试,但是跟我原本猜想的好像不太一样:
测试过程我把 encoder predictor decoder 分别导出为 onnx-gpu模型,然后单独测试其耗时; 1. 测试结果【测试音频不同】2. 测试代码:
3. decoder onnx 导出代码
|
Is this issue fixed? |
0. 【问题】[paraformer]When is ONNX GPU export supported?
1. 版本**【wenet-v3.0.1】**
2. 尝试对 paraformer onnx gpu 进行转换
The text was updated successfully, but these errors were encountered: