ArtemisZGL

test,py get different result with train,py

5

when i load the checkpoint to use test.py to test the valid set, it get very higher wer and cer, which is much higher than valid cer showing in training....

[Question] what is the meaning of src,tgt len？

1

i found the src max len is fist use in the dataloader for spec, and if the sample rate is 16K, win size is 320, hop size is 160, then...

i use librispeech 100h to train the model with 6 2080ti. But i found the gpu usage is very low (about 20% with batch size 24). And i have use...

``` neighbors = self.neighbor_prefixes[prefix] neighbors.remove(uttname) ``` I found the code in getitem function in LibriSpeechSegTupleWavDataset just remove the uttname in the neighbors, and it will remove it in the self.neighbor_prefixes...

how to set the param when converting ? I don't have the trans model but all in the mdl, and what does config file mean or module mean? ``` Traceback...

First thanks to you great work ! But I have some questions about it, why in the training code doesn't have validation part ? Can I just add it like...

according the link ( http://fluent.ai:2052/jf8398hf30f0381738rucj3828chfdnchs.tar.gz ) given in the paper and the fluent.ai, finally I found the link is not working, showing 521. Maybe the server is not working anymore.

about the dependency

2

could you share the dependency for this repo?

❓ [Question] a10 performance drop significantly

2

## ❓ Question I converted the gfpgan model (https://github.com/TencentARC/GFPGAN) with torch_tensorrt, and I found torch_tensorrt is twice as fast as torch in 3070. But in one a10 server, torch_tensorrt and...

how many data should i use to train this model？

4

Thanks for your job first, I want to know that how many data should i use to train in your repo. I want to use the cmu arctic dataset for...