trunglebka

First: Yes, I tried unchecking validate certificates. But since my OS is the recent Fedora version (36) with latest updated packages and google is a very popular website I don't...

i have same problem

I've encontered this problem too, using a stopwatch it is about 3.30 seconds (52 lines of code) with ssh remote. The CPU is intel E5-2695 V2 Just install black from...

For this question, I've tried both cuda decoder v1 vs v2 and v1 give better RTF in my case so my old service using this implementation. Maybe nvidia provided kaldi...

I've retired from my old company. In my case, after some experiment of tuning parameters using nvidia kaldi docker with T4, v1 give me about 500 RTFx but v2 just...

Yeah, just want to provide you context where v1 being used.

@desh2608 No, I didn't run it. Just run [validate_manifest.py](https://github.com/k2-fsa/icefall/blob/master/egs/librispeech/ASR/local/validate_manifest.py)

@desh2608 I tried `fix_manifests`, but it still happens. Sorry for the delayed reply, but it's a huge dataset, so it takes time.

@ Here is the code that I used to prepare the data. I removed the data dir (data/nsc) and re-ran [prepare-nsc.sh](https://gist.github.com/trunglebka/725ec0a8257b9369ac9f9a52d9a862b3#file-prepare-nsc-sh) from the beginning Not sure if I am using...

@pzelasko Using your suggestion, I encountered this error: ``` File "/opt/anaconda3/envs/k2/lib/python3.11/site-packages/lhotse/dataset/speech_recognition.py", line 99, in __getitem__ validate_for_asr(cuts) File "/opt/anaconda3/envs/k2/lib/python3.11/site-packages/lhotse/dataset/speech_recognition.py", line 218, in validate_for_asr assert supervision.end