Skip to content

Latest commit

 

History

History

transducer_stateless_modified

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
README.md
README.md
 
 
__init__.py
__init__.py
 
 
asr_datamodule.py
asr_datamodule.py
 
 
beam_search.py
beam_search.py
 
 
conformer.py
conformer.py
 
 
decode.py
decode.py
 
 
decoder.py
decoder.py
 
 
encoder_interface.py
encoder_interface.py
 
 
export.py
export.py
 
 
joiner.py
joiner.py
 
 
model.py
model.py
 
 
pretrained.py
pretrained.py
 
 
subsampling.py
subsampling.py
 
 
test_decoder.py
test_decoder.py
 
 
train.py
train.py
 
 
transformer.py
transformer.py
 
 

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/aishell/ASR

export CUDA_VISIBLE_DEVICES="0,1,2,3,4,5,6,7"

./transducer_stateless_modified/train.py \
  --world-size 8 \
  --num-epochs 30 \
  --start-epoch 0 \
  --exp-dir transducer_stateless_modified/exp \
  --max-duration 250 \
  --lr-factor 2.5