mirrors/icefall
1
0
Fork 0
mirror of https://github.com/k2-fsa/icefall.git synced 2025-08-27 10:44:19 +00:00
Code Issues Packages Projects Releases Wiki Activity
icefall/egs/librispeech/ASR/transducer
History
Fangjun Kuang d362a3dba7 Reduce the number of decoder layers from 4 to 2.
2021-12-18 11:25:33 +08:00
..
__init__.py
Apply layer normalization to the output of each gate in LSTM/GRU. (#139)
2021-12-07 18:38:03 +08:00
asr_datamodule.py
RNN-T Conformer training for LibriSpeech (#143)
2021-12-18 07:42:51 +08:00
beam_search.py
Remove sos ID.
2021-12-18 11:22:12 +08:00
conformer.py
Replace BatchNorm in the Conformer model with LayerNorm.
2021-12-18 11:12:38 +08:00
decode.py
Reduce the number of decoder layers from 4 to 2.
2021-12-18 11:25:33 +08:00
decoder.py
Remove sos ID.
2021-12-18 11:22:12 +08:00
encoder_interface.py
RNN-T Conformer training for LibriSpeech (#143)
2021-12-18 07:42:51 +08:00
export.py
Reduce the number of decoder layers from 4 to 2.
2021-12-18 11:25:33 +08:00
joiner.py
Use tanh in the joint network.
2021-12-18 11:14:05 +08:00
model.py
Remove sos ID.
2021-12-18 11:22:12 +08:00
pretrained.py
Reduce the number of decoder layers from 4 to 2.
2021-12-18 11:25:33 +08:00
README.md
RNN-T Conformer training for LibriSpeech (#143)
2021-12-18 07:42:51 +08:00
rnn.py
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_conformer.py
Remove sos ID.
2021-12-18 11:22:12 +08:00
test_decoder.py
Remove sos ID.
2021-12-18 11:22:12 +08:00
test_joiner.py
RNN-T Conformer training for LibriSpeech (#143)
2021-12-18 07:42:51 +08:00
test_rnn.py
RNN-T Conformer training for LibriSpeech (#143)
2021-12-18 07:42:51 +08:00
test_transducer.py
Remove sos ID.
2021-12-18 11:22:12 +08:00
test_transformer.py
Remove sos ID.
2021-12-18 11:22:12 +08:00
train.py
Reduce the number of decoder layers from 4 to 2.
2021-12-18 11:25:33 +08:00
transformer.py
Remove input feature batchnorm..
2021-12-18 11:05:28 +08:00

README.md

Introduction

The encoder consists of Conformer layers in this folder. You can use the following command to start the training:

cd egs/librispeech/ASR

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

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