yangsuxia

> Hey @yangsuxia - you should use `return_timestamps=True` in this case as it greatly reduces model hallucinations. Regarding benchmarks, see #5. Yes, after setting the parameter return_timestamps to true, the...

> > the other is to insert multiple words in the middle > > This problem is much harder to fix in your case since the ground truth contains 3...

> > one is to delete words at the end, > > Could you add some tail paddings to your utterances and retry? I added 1s silence after the testset,...

> What is your original setting? The above reply contains the original configuration for decoding. It was written incorrectly. The avg used to be 2, but now it is 1

Hi, what is the final loss value when you are in finetune(10h data). Looking forward to your reply, thank you.

Has the problem been resolved? I also encountered the same problem

> hi @starshine360 , thank you for your bug report. Could you please send the exact command line you run? Thanks. I run it like this： zcat seamless.dataset.metadata.public.arb-enA.tsv.gz |egrep "^crawl"|tr...

解决方案具体怎么做啊？没看懂，ctc和decoder怎么使用不同的tokenizer > 正常现象，解决方案是 #2247