Huang Kuan Po

Results 9 comments of Huang Kuan Po

Thanks for your reply, but I still hope that the asr model forwards with GPU calculation.

I left [COMPUTE_METRICS](https://github.com/asteroid-team/asteroid/blob/master/egs/librimix/ConvTasNet/eval.py#L50) as an empty list. However, it is still running so slow. ``` 0%| | 2/3000 [03:17

I changed line 231 in [metrics.py](https://github.com/asteroid-team/asteroid/blob/master/asteroid/metrics.py#L231) into ```self.asr_model = Speech2Text(**d.download_and_unpack(model_name), device='cuda')``` and added ```wav = torch.from_numpy(wav).cuda()``` to the ```predict_hypothesis``` function between [line 344 and 345](https://github.com/asteroid-team/asteroid/blob/master/asteroid/metrics.py#L344). It is now able to...

@popcornell @mpariente Thanks for your replies. I think the most urgent thing we need now is batch processing on asr. I think this requires a lot of modification in [eval.py](https://github.com/asteroid-team/asteroid/blob/master/egs/librimix/ConvTasNet/eval.py)....

@popcornell I have sent a PR allowing asr to compute with GPU.

Do we still have to provide the `model.w2v_path` parameter in the command?

should be `out.update({"fad_score":fad_score})`