README.md

# Introduction

This repo contains pre-trained model using
<https://github.com/k2-fsa/icefall/pull/213>.

It is trained on train-clean-100 subset of the LibriSpeech dataset.
Also, it uses the `S` subset from GigaSpeech as extra training data.

## How to clone this repo
```
sudo apt-get install git-lfs
git clone https://huggingface.co/csukuangfj/icefall-asr-librispeech-100h-transducer-stateless-multi-datasets-bpe-500-2022-02-21

cd icefall-asr-librispeech-100h-transducer-stateless-multi-datasets-bpe-500-2022-02-21
git lfs pull
```

**Catuion**: You have to run `git lfs pull`. Otherwise, you will be SAD later.

The model in this repo is trained using the commit `2332ba312d7ce72f08c7bac1e3312f7e3dd722dc`.

You can use

```
git clone https://github.com/k2-fsa/icefall
cd icefall
git checkout 2332ba312d7ce72f08c7bac1e3312f7e3dd722dc
```
to download `icefall`.

You can find the model information by visiting <https://github.com/k2-fsa/icefall/blob/2332ba312d7ce72f08c7bac1e3312f7e3dd722dc/egs/librispeech/ASR/transducer_stateless_multi_datasets/train.py#L198>.

In short, the encoder is a Conformer model with 8 heads, 12 encoder layers, 512-dim attention, 2048-dim feedforward;
the decoder contains a 1024-dim embedding layer and a Conv1d with kernel size 2.

The decoder architecture is modified from
[Rnn-Transducer with Stateless Prediction Network](https://ieeexplore.ieee.org/document/9054419).
A Conv1d layer is placed right after the input embedding layer.

-----

## Description

This repo provides pre-trained transducer Conformer model for the LibriSpeech dataset
using [icefall][icefall]. There are no RNNs in the decoder. The decoder is stateless
and contains only an embedding layer and a Conv1d.

The commands for training are:

```
cd egs/librispeech/ASR/
./prepare.sh
./prepare_giga_speech.sh

export CUDA_VISIBLE_DEVICES="0,1"

./transducer_stateless_multi_datasets/train.py \
  --world-size 2 \
  --num-epochs 60 \
  --start-epoch 0 \
  --exp-dir transducer_stateless_multi_datasets/exp-100-2 \
  --full-libri 0 \
  --max-duration 300 \
  --lr-factor 1 \
  --bpe-model data/lang_bpe_500/bpe.model \
  --modified-transducer-prob 0.25
  --giga-prob 0.2
```

The tensorboard training log can be found at
<https://tensorboard.dev/experiment/qUEKzMnrTZmOz1EXPda9RA/>

The command for decoding is:
```
epoch=57
avg=17

## greedy search
for epoch in 57; do
  for avg in 17; do
    for sym in 1 2 3; do
    ./transducer_stateless_multi_datasets/decode.py \
      --epoch $epoch \
      --avg $avg \
      --exp-dir transducer_stateless_multi_datasets/exp-100-2 \
      --bpe-model ./data/lang_bpe_500/bpe.model \
      --max-duration 100 \
      --context-size 2 \
      --max-sym-per-frame $sym
    done
  done
done

## modified beam search

epoch=57
avg=17
./transducer_stateless_multi_datasets/decode.py \
  --epoch $epoch \
  --avg $avg \
  --exp-dir transducer_stateless_multi_datasets/exp-100-2 \
  --bpe-model ./data/lang_bpe_500/bpe.model \
  --max-duration 100 \
  --context-size 2 \
  --decoding-method modified_beam_search \
  --beam-size 4
```

You can find the decoding log for the above command in this
repo (in the folder `log`).

The WERs for the test datasets are

|                                     | test-clean | test-other | comment                                  |
|-------------------------------------|------------|------------|------------------------------------------|
| greedy search (max sym per frame 1) | 6.34       | 16.7       | --epoch 57, --avg 17, --max-duration 100 |
| greedy search (max sym per frame 2) | 6.34       | 16.7       | --epoch 57, --avg 17, --max-duration 100 |
| greedy search (max sym per frame 3) | 6.34       | 16.7       | --epoch 57, --avg 17, --max-duration 100 |
| modified beam search (beam size 4)  | 6.31       | 16.3       | --epoch 57, --avg 17, --max-duration 100 |


# File description

- [log][log], this directory contains the decoding log and decoding results
- [test_wavs][test_wavs], this directory contains wave files for testing the pre-trained model
- [data][data], this directory contains files generated by [prepare.sh][prepare]
- [exp][exp], this directory contains only one file: `preprained.pt`

`exp/pretrained.pt` is generated by the following command:

```bash
./transducer_stateless_multi_datasets/export.py \
  --epoch 57 \
  --avg 17 \
  --bpe-model data/lang_bpe_500/bpe.model \
  --exp-dir transducer_stateless_multi_datasets/exp-full
```

**HINT**: To use `pretrained.pt` to compute the WER for test-clean and test-other,
just do the following:
```
cp icefall-asr-librispeech-100h-transducer-stateless-multi-datasets-bpe-500-2022-02-21/exp/pretrained.pt \
  /path/to/icefall/egs/librispeech/ASR/transducer_stateless_multi_datasets/exp/epoch-999.pt
```
and pass `--epoch 999 --avg 1` to `transducer_stateless_multi_datasets/decode.py`.


[icefall]: https://github.com/k2-fsa/icefall
[prepare]: https://github.com/k2-fsa/icefall/blob/master/egs/librispeech/ASR/prepare.sh
[exp]: https://huggingface.co/csukuangfj/icefall-asr-librispeech-100h-transducer-stateless-multi-datasets-bpe-500-2022-02-21/tree/main/exp
[data]: https://huggingface.co/csukuangfj/icefall-asr-librispeech-100h-transducer-stateless-multi-datasets-bpe-500-2022-02-21/tree/main/data
[test_wavs]: https://huggingface.co/csukuangfj/icefall-asr-librispeech-100h-transducer-stateless-multi-datasets-bpe-500-2022-02-21/tree/main/test_wavs
[log]: https://huggingface.co/csukuangfj/icefall-asr-librispeech-100h-transducer-stateless-multi-datasets-bpe-500-2022-02-21/tree/main/log
[icefall]: https://github.com/k2-fsa/icefall