Shashi Kant

Hey thanks, I will correct that

Thanks will update this in next release.

Thanks for the awesome contributions from you two! Adding some of my minor observations relevant to this: 1. **Layer norm issue:** +1, I observed similar behavior. 2. **`bert_attention_plugin` weird behavior:**...

Thank you so much @symphonylyh for the guidelines! > Lastly, from a practical perspective, w/o BERT plugin path has a limitation on padding removal -- that is, when you have...

> @shashikg We actually could remove the padding 30s restriction of encoder, see https://github.com/k2-fsa/icefall/blob/master/egs/aishell/ASR/whisper/whisper_encoder_forward_monkey_patch.py#L15. It would save cross kv cache VRAM usage as well. Hey yes, I agree and most...

1. Hey I think HF ChunkPipeline sets anything greater than `num_worker>0` to `num_worker=1`. See [here](https://github.com/huggingface/transformers/blob/main/src/transformers/pipelines/base.py#L1173). Though I will once run the benchmark after setting this to a higher number. 2....

On CPU recommend to use any smaller model: `model = whisper_s2t.load_model("tiny", device='cpu', compute_type='float32')`

@twardoch this is a bug for the aligner model. By default for alignment tiny model is used which expects n_mels to be of size 80 but large-v3 expects n_mels to...

Hi @Pranjalya @Sembiance ! Can you describe here or link an issue related to small duration audio?

Hey @BBC-Esq ! I think there can be a simple fix for this. I will add the fix in next release. PS: I'm slightly stuffed with my office work. Expect...