revdotcom
config.yaml
Hi, In this inference code, where is config.yaml file ?
python3 wenet/bin/recognize_wav.py --config config.yaml --checkpoint /home/Ubuntu/myaudio/reverb_asr_v1.pt --audio /home/Ubuntu/myaudio/jfk.wav --result_dir /home/Ubuntu/myaudio/output
In addition to above, I am also getting this error :
(rev) Ubuntu@0041-kci-prxmx10171:~/reverb/asr$ python3 wenet/bin/recognize_wav.py --checkpoint /home/Ubuntu/myaudio/reverb_asr_v1.pt --audio /home/Ubuntu/myaudio/jfk.wav --result_dir /home/Ubuntu/myaudio/output
Traceback (most recent call last):
File "/home/Ubuntu/reverb/asr/wenet/bin/recognize_wav.py", line 32, in
(rev) Ubuntu@0041-kci-prxmx10171:~/reverb/asr$ ls app.py pyproject.toml README.md requirements.txt wenet wer_evaluation
The config.yaml file can be found along the checkpoint file on HF : https://huggingface.co/Revai/reverb-asr/tree/main
Did you download the whole folder content ? You will need the other files in there, like the tokenizer (tk.*) and the en-cmvn.json.
The 2nd issue is likely because you forgot this part of the instructions:
export PYTHONPATH="$(pwd)"/asr:$PYTHONPATH # adding this to make wenet/ work
Hi, I have followed your repo instructions (which I am sorry to say, very badly written). Its still not working . I have already spent like 3 hours on getting ti working. I am pasting all of steps and the error. If you are open to feedback then allow me to say that if you guys are launching a product to OpenAI's whisper, please at least make installation simple or to the very least installation instructions clear enough :
===============
` conda create -n reverb-env python=3.10 -y && conda activate reverb-env
git clone https://github.com/revdotcom/reverb.git && cd reverb
pip install -r asr/requirements.txt pip install -r diarization/requirements.txt export PYTHONPATH="$(pwd)"/asr:$PYTHONPATH
pip install huggingface_hub
huggingface-cli login
git lfs install
huggingface-cli download --resume-download Revai/reverb-asr --local-dir checkpoints --local-dir-use-symlinks False
pip install git+https://github.com/wenet-e2e/wenet.git and also tried with git clone wenet
(reverb-env) Ubuntu@0041-kci-prxmx10171:~/reverb$ python wenet/bin/recognize_wav.py --config /home/Ubuntu/reverb/checkpoints/config.yaml --checkpoint /home/Ubuntu/reverb/checkpoints/reverb_asr_v1.pt --audio /home/Ubuntu/myaudio/jfk.wav --result_dir /home/Ubuntu/reverb/output --modes ctc_prefix_beam_search attention_rescoring --gpu 0 --verbatimicity 1.0 python: can't open file '/home/Ubuntu/reverb/wenet/bin/recognize_wav.py': [Errno 2] No such file or directory (reverb-env) Ubuntu@0041-kci-prxmx10171:~/reverb$ (reverb-env) Ubuntu@0041-kci-prxmx10171:~/reverb$ (reverb-env) Ubuntu@0041-kci-prxmx10171:~/reverb$ ls -ltr total 40 -rw-rw-r-- 1 Ubuntu Ubuntu 3814 Oct 4 21:48 README.md -rw-rw-r-- 1 Ubuntu Ubuntu 11112 Oct 4 21:48 LICENSE -rw-rw-r-- 1 Ubuntu Ubuntu 1081 Oct 4 21:48 Dockerfile drwxrwxr-x 2 Ubuntu Ubuntu 4096 Oct 4 21:48 resources drwxrwxr-x 3 Ubuntu Ubuntu 4096 Oct 4 21:48 diarization drwxrwxr-x 4 Ubuntu Ubuntu 4096 Oct 4 21:48 asr drwxrwxr-x 3 Ubuntu Ubuntu 4096 Oct 4 21:53 checkpoints drwxrwxr-x 2 Ubuntu Ubuntu 4096 Oct 4 21:55 output
PS. I have also tried above after git cloning wnet directory from their repo, but still doesnt work.
`