YAML Metadata
Warning:
empty or missing yaml metadata in repo card
(https://huggingface.co/docs/hub/model-cards#model-card-metadata)
Instructions for ESPNet Inference Environment Creation
- Start with the the espnet docker image
espnet/espnet:gpu-latest
- Create a container with the required resources. It is essential that you increase the default shared memory to 8GB with
--shm-size 8GB
- Inside the docker container, do
pip install espnet
- Next, open file
/opt/miniconda/lib/python3.9/site-packages/espnet2/asr/espnet_model.py
and navigate to line121
and change it fromfrom warprnnt_pytorch import RNNTLoss
tofrom torchaudio.transforms import RNNTLoss
and go to line 125 and remove it.
123 self.criterion_transducer = RNNTLoss(
124 blank=self.blank_id,
125 fastemit_lambda=0.0, # ---> Remove this line
126 )
Now you are ready to run the inference.py script. To run inference
python inference.py --config branchformer/config.yaml --model_file branchformer/40epoch.pth --audio test.wav
You might have to update paths in the branchformer/config.yaml as per your requirements for your changes.
Inference Providers
NEW
This model isn't deployed by any Inference Provider.
๐
Ask for provider support