icefall

icefall copied to clipboard

JinZr

Severe Deletion Errors with pruned_transducer_stateless7_streaming on Japanese Dataset Validation

6

Hi Next-gen Kaldi team, Thank you for your detailed documentation and support through the WeChat discussion group. We have been developing a new recipe for our open-sourced Japanese corpus, `ReazonSpeech`,...

Triplecq

Zipformer streaming model: repeated single-syllable words/characters stop abruptly

3

I have been experimenting with the pre-trained zipformer streaming models: * https://huggingface.co/marcoyang/icefall-libri-giga-pruned-transducer-stateless7-streaming-2023-04-04 * https://huggingface.co/pkufool/icefall-asr-zipformer-streaming-wenetspeech-20230615 A problem I observe is the models do not recognise more than two single-syllable words/characters in...

rouseabout

Fast Text-Only Domain Adaptation of RNN-Transducer Prediction Network

5

Hello guys, In this article https://arxiv.org/pdf/2104.11127.pdf there is very simple idea of how to adapt RNN-T to the unknown text. The writer of the article reports about very nice WER...

AlexandderGorodetski

Failed in reproduce result in PromptASR

2

Hi everyone, I've downloaded the pre-trained models of PromptASR to reproduce the result that has been report. However, when I reproduce with the recipes, the result between using prompt (use_pre_text=True)...

tienanh28122000

JinZr

Use padding_idx=None for nn.Embedding() in the decoder model

2

We need to change sherpa/sherpa-onnx/sherpa-ncnn to use `[-1, 0]` as the initial tokens during decoding instead of `[0, 0]`. The consequence is that if we need to re-export the model,...

csukuangfj

Fix Blankskip bug.

3

kobenaxie

further refinement to the swbd recipe

4

including a narrowband setup

JinZr

Streaming Zipformer2. extra words at the end.

6

Hello, I tried training a streaming zipformer2, but I end up with small common words (like "and" or "it" or letters randomly being added to the output at the end...

joazoa