2021-12-17 20:39:34 +08:00

20 lines
379 B
Markdown

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