icefall
icefall copied to clipboard
k2-fsa
When recognizing, if a sentence is completely silent, blank skip will reduce all frame. Exception occurs: Traceback (most recent call last): File "/data/k2/icefall/egs/xxxx/Rework/./pruned_transducer_stateless7_ctc_bs/ctc_guide_decode_bs.py", line 847, in main() File "/data/k2/miniconda3/envs/k2-1080ti/lib/python3.9/site-packages/torch/autograd/grad_mode.py", line...
Added zipformer ctc streaming support for librispeech at egs/librispeech/ASR/zipformer_ctc_streaming
Added zipformer ctc streaming support for librispeech (egs/librispeech/ASR/zipformer_ctc_streaming) Pls refer https://github.com/k2-fsa/icefall/issues/999#issuecomment-1569761925
Could you please add support for [streaming model export to jit trace format](https://github.com/k2-fsa/icefall/blob/master/egs/librispeech/ASR/conformer_ctc3/export.py#L267)?
https://github.com/k2-fsa/icefall/pull/1058 adds updated zipformer. However, we only have `torch.jit.script()` support for it at present. We need to export it to onnx so that it can be used in [sherpa-onnx](https://github.com/k2-fsa/sherpa-onnx) Help...
According to the link shown below, https://huggingface.co/csukuangfj/sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20 How to train a bilingual Chinese-English zipformer model from icefall/egs/librispeech/ASR/pruned_transducer_stateless7 _streaming? Because the data preparation part is based on English speech corpus LibriSpeech，there...
When importing data from Kaldi, lhotse requires that there must exist a `segments` file or a `utt2spk` file in order to generate a supervision file. This PR provides a standalone...
**This PR is not for merge**. We only provide detailed commit history of the upgraded Zipformer here.
Hi, I integrate the codes of Zipformer as well as ScaleAdam/Eden Optimizer to my projects and reach satisfactory results when I train the model using 8 GPUs on a single...
This PR is **not** for merge. It's only for illustration purposes. This PR adds cross-attention to zipformer.
