mirrors/icefall
1
0
Fork 0
mirror of https://github.com/k2-fsa/icefall.git synced 2025-08-09 18:12:19 +00:00
Code Issues Packages Projects Releases Wiki Activity
icefall/egs/librispeech/ASR/RESULTS.md
pkufool ef233486ae
The training script produce WER of 2.57% on librispeech test-clean (#13) 
* Add grad_clip and weight-decay, small fix of dataloader and masking

* Add RESULTS.md
2021-08-20 10:08:08 +08:00

936 B
Raw Blame History

Results

LibriSpeech BPE training results (Conformer-CTC)

2021-08-19

(Wei Kang): Result of https://github.com/k2-fsa/icefall/pull/13

TensorBoard log is available at https://tensorboard.dev/experiment/GnRzq8WWQW62dK4bklXBTg/#scalars

Pretrained model is available at https://huggingface.co/pkufool/conformer_ctc

The best decoding results (WER) are listed below, we got this results by averaging models from epoch 15 to 34, and using attention-decoder decoder with num_paths equals to 100.

test-clean test-other
WER 2.57% 5.94%

To get more unique paths, we scaled the lattice.scores with 0.5 (see https://github.com/k2-fsa/icefall/pull/10#discussion_r690951662 for more details), we searched the lm_score_scale and attention_score_scale for best results, the scales that produced the WER above are also listed below.

lm_scale attention_scale
test-clean 1.3 1.2
test-other 1.2 1.1