Masao Someki
Masao Someki
## What? Bugfix for #5614. See the PR for more detail. ## See also #5614
## What? This PR is to support several external dataset libraries or formats and create instruction documents. Currently, I'm planning to support the following datasets: - [x] torchdata (torchaudio.dataset) -...
**Describe the bug** Using the pre-trained models causes unnecessary GPU memory allocation. It seems there might be inefficiencies in the allocation process. It would be beneficial to optimize this behavior...
## What? Support partially auto-regressive decoding for ASR and S2T task. Since this decoding process is based on BERT-CTC, we can only perform speech recognition. ## Why? To speed up...
## What? Adds integration tests for ESPnet EZ - [x] ASR1 - [x] basic - [x] streaming - [x] RNN-Transducer - [x] ASR Transducer (standalone) - [x] ASR2 - [x]...