mirrors/icefall
1
0
Fork 0
mirror of https://github.com/k2-fsa/icefall.git synced 2025-08-08 09:32:20 +00:00
Code Issues Packages Projects Releases Wiki Activity
icefall/egs/librispeech/ASR/transducer_stateless
History
Fangjun Kuang a8150021e0
Use modified transducer loss in training. (#179) 
* Use modified transducer loss in training.

* Minor fix.

* Add modified beam search.

* Add modified beam search.

* Minor fixes.

* Fix typo.

* Update RESULTS.

* Fix a typo.

* Minor fixes.
2022-02-07 18:37:36 +08:00
..
__init__.py
RNN-T Conformer training for LibriSpeech (#143)
2021-12-18 07:42:51 +08:00
asr_datamodule.py
RNN-T Conformer training for LibriSpeech (#143)
2021-12-18 07:42:51 +08:00
beam_search.py
Use modified transducer loss in training. (#179)
2022-02-07 18:37:36 +08:00
conformer.py
Remove batchnorm, weight decay, and SOS from transducer conformer encoder (#155)
2021-12-27 16:01:10 +08:00
decode.py
Use modified transducer loss in training. (#179)
2022-02-07 18:37:36 +08:00
decoder.py
Use modified transducer loss in training. (#179)
2022-02-07 18:37:36 +08:00
encoder_interface.py
RNN-T Conformer training for LibriSpeech (#143)
2021-12-18 07:42:51 +08:00
export.py
Minor fixes (#193)
2022-01-27 18:01:17 +08:00
joiner.py
Use optimized_transducer to compute transducer loss. (#162)
2022-01-10 11:54:58 +08:00
model.py
Use modified transducer loss in training. (#179)
2022-02-07 18:37:36 +08:00
pretrained.py
Use modified transducer loss in training. (#179)
2022-02-07 18:37:36 +08:00
README.md
RNN-T Conformer training for LibriSpeech (#143)
2021-12-18 07:42:51 +08:00
subsampling.py
RNN-T Conformer training for LibriSpeech (#143)
2021-12-18 07:42:51 +08:00
test_decoder.py
Increase the size of the context in the RNN-T decoder. (#153)
2021-12-23 07:55:02 +08:00
train.py
Use modified transducer loss in training. (#179)
2022-02-07 18:37:36 +08:00
transformer.py
Remove batchnorm, weight decay, and SOS from transducer conformer encoder (#155)
2021-12-27 16:01:10 +08:00

README.md

Introduction

The decoder, i.e., the prediction network, is from https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=9054419 (Rnn-Transducer with Stateless Prediction Network)

You can use the following command to start the training:

cd egs/librispeech/ASR

export CUDA_VISIBLE_DEVICES="0,1,2,3"

./transducer_stateless/train.py \
  --world-size 4 \
  --num-epochs 30 \
  --start-epoch 0 \
  --exp-dir transducer_stateless/exp \
  --full-libri 1 \
  --max-duration 250 \
  --lr-factor 2.5