agent404's picture
upload files
fe781a6
# RepCodec: A Speech Representation Codec for Speech Tokenization
> [**RepCodec: A Speech Representation Codec for Speech Tokenization**](https://arxiv.org/abs/2309.00169)
## Introduction
**RepCodec** is a speech tokenization method for converting a speech waveform into a sequence of discrete semantic
tokens.
The main idea is to train a representation codec which learns a vector quantization codebook through reconstructing the
input speech representations from speech encoders like HuBERT or data2vec.
Extensive experiments show that RepCodec significantly outperforms the widely used k-means clustering approach in both
speech understanding and generation.
Also, RepCodec generalizes well across various speech encoders and languages.
<img src="images/RepCodec.png" alt="se" width="1000" />
## RepCodec Models
| Feature Type | Speech Data | RepCodec Model |
|-----------------------------------------------------------------------------------------------------------------------|----------------------------------------------------------|----------------------------------------------------------------------------------------------------------|
| [HuBERT base](https://github.com/facebookresearch/fairseq/tree/main/examples/hubert#pre-trained-and-fine-tuned-asr-models) layer 9 | [Librispeech](http://www.openslr.org/12) train-clean-100 | [hubert_base_l9](https://drive.google.com/file/d/1XD0HKl607FFjri2-VJT7lHQeSpxsCCFO/view?usp=sharing) |
| [HuBERT large](https://github.com/facebookresearch/fairseq/tree/main/examples/hubert#pre-trained-and-fine-tuned-asr-models) layer 18 | [Librispeech](http://www.openslr.org/12) train-clean-100 | [hubert_large_l18](https://drive.google.com/file/d/1mTbm5GeJ7gp_5L3QLP-JGXdf8RnRw5n6/view?usp=sharing) |
| [data2vec base](https://github.com/facebookresearch/fairseq/blob/main/examples/data2vec/README.md#speech-2) layer 6 | [Librispeech](http://www.openslr.org/12) train-clean-100 | [data2vec_base_l6](https://drive.google.com/file/d/1d8sf3Ko_fYM9zlaiwxK_4xusLRKV5EMd/view?usp=sharing) |
| [data2vec large](https://github.com/facebookresearch/fairseq/blob/main/examples/data2vec/README.md#speech-2) layer 18 | [Librispeech](http://www.openslr.org/12) train-clean-100 | [data2vec_large_l18](https://drive.google.com/file/d/1nuRIHaejT-uVi4cluftbT8o_JZqar5SU/view?usp=sharing) |
| [Whisper medium](https://github.com/openai/whisper/tree/main#available-models-and-languages) layer 24 | [Librispeech](http://www.openslr.org/12) train-clean-100 | [whisper_medium_l24](https://drive.google.com/file/d/1V6YJSA2V4iywXrecJAN0oqsa3aHowexZ/view?usp=sharing) |
| [Whisper large-v2](https://github.com/openai/whisper/tree/main#available-models-and-languages) layer 32 | [Librispeech](http://www.openslr.org/12) train-clean-100 | [whisper_large_l32](https://drive.google.com/file/d/1k_X7ZMPg8iOeDrIJe70v6CHfFygzufXC/view?usp=sharing) |
## Speech Tokenization Using Pre-Trained Models
### Installation
Please first install RepCodec by
```
git clone https://github.com/mct10/RepCodec.git
cd RepCodec
pip install .
```
We used Python 3.9.18 and PyTorch 1.12.1 to test the usage, but the code should be compatible with other recent Python
and PyTorch versions.
### Representation Preparation
We adapt the `dump_hubert_feature.py` script
from [fairseq](https://github.com/facebookresearch/fairseq/tree/main/examples/hubert/simple_kmeans#hubert-feature)
to support dumping representations from **data2vec**, **HuBERT**, or **Whisper** encoders.
If you use our script (`examples/dump_feature.py`), please also install the following packages:
```
pip install npy_append_array soundfile
```
Additionally, if you want to dump representations from
- **data2vec** or **HuBERT**: please
follow [fairseq's instruction](https://github.com/facebookresearch/fairseq#requirements-and-installation) to install
the latest fairseq.
- **Whisper**: please follow [Whispers'instruction](https://github.com/openai/whisper/tree/main#setup) to install the
latest
Whisper.
Then, you can follow the given examples to dump representations:
```
# Example 1: dump from HuBERT base layer 9
# (for data2vec, simply change "model_type" to data2vec and "ckpt_path" to the path of data2vec model)
layer=9
python3 examples/dump_feature.py \
--model_type hubert \
--tsv_path /path/to/tsv/file \
--ckpt_path /path/to/HuBERT/model \
--layer ${layer} \
--feat_dir /dir/to/save/representations
# Example 2: dump from Whisper medium layer 24
layer=24
python3 examples/dump_feature.py \
--model_type whisper \
--tsv_path /path/to/tsv/file \
--whisper_root /directory/to/save/whisper/model \
--whisper_name medium \
--layer ${layer} \
--feat_dir /dir/to/save/representations
```
Explanations about the args:
- **model_type:** choose from `data2vec`, `hubert`, and `whisper`.
- **tsv_path:** path of the tsv file.
Should have the format of
```
/dir/to/dataset
path_of_utterance_1 number_of_frames
path_of_utterance_2 number_of_frames
```
You can follow [this script](https://github.com/facebookresearch/fairseq/blob/main/examples/wav2vec/wav2vec_manifest.py)
to generate the tsv file.
For example, by running
```
python wav2vec_manifest.py \
/dir/to/LibriSpeech/dev-clean \
--dest /dir/to/manifest \
--ext flac \
--valid-percent 0
```
you can obtain the `dev-clean.tsv` in `/dir/to/manifest` for LibriSpeech. (By default, the output file name
is `train.tsv`. Remember to rename the file.)
It should be similar to:
```
/dir/to/LibriSpeech/dev-clean
2277/149896/2277-149896-0026.flac 78720
2277/149896/2277-149896-0005.flac 89600
2277/149896/2277-149896-0033.flac 45520
```
- **ckpt_path**:
must provide for data2vec and HuBERT.
You need to download the model
from [data2vec website](https://github.com/facebookresearch/fairseq/blob/main/examples/data2vec/README.md#speech-2)
or [HuBERT website](https://github.com/facebookresearch/fairseq/tree/main/examples/hubert#pre-trained-and-fine-tuned-asr-models)
yourself.
`--ckpt_path` is the path of the data2vec/HuBERT model.
- **whisper_root** and **whisper_name**:
must provide **BOTH** `--whisper_root` and `--whisper_name` for Whisper.
If there is no corresponding model in `--whisper_root`, the script will download for you.
- **layer**:
which Transformer encoder layer of the model should the representations be extracted from.
It is **1-based**.
For example, if layer=9, then the outputs from the 9<sup>th</sup> Transformer encoder layer are dumped.
Range: [1, number of Transformer encoder layers]
- **feat_dir**: The output representations will be saved to `${feat_dir}/0_1.npy`
and `${feat_dir}/0_1.len`.
For other useful functionalities (e.g., sharding), please check the argument list in `examples/dump_feature.py`.
### Command Line Usage
We expect to have `${feat_dir}/0_1.npy` and `${feat_dir}/0_1.len` in the provided
directory `/dir/to/representaitons`.
Also, the tsv file should be the **same** as the one used in [Representation Preparation](#representation-preparation).
```
repcodec /dir/to/representaitons \
--model /path/to/repcodec/model \
--tsv_path /path/to/tsv/file \
[--model_config_path /path/to/train/config] \
[--use_gpu] \
[--out_dir /path/to/output]
```
If you trained the model yourself following [Training New RepCodec Models](#training-new-repcodec-models),
please provide the training config file using `--model_config_path`.
If you use the model we provide [here](#repcodec-models), then you do not have to provide that.
This command will tokenize the representations and the output discrete tokens will be saved to `${out_dir}/tokens`.
The tokens are in the same order as the provided tsv file.
An example of the output file:
```
/dir/to/LibriSpeech/dev-clean
2277/149896/2277-149896-0026.flac 696 696 198 198 198 498 ...
2277/149896/2277-149896-0005.flac 696 696 198 198 198 907 ...
2277/149896/2277-149896-0033.flac 696 696 198 198 198 696 ...
```
Under `examples/tokens`, we provide some token files as references. They are obtained from LibriSpeech dev-clean subset
using the 6 types of representations and corresponding [RepCodec Models](#repcodec-models).
Your results should be very similar to ours.
### Python Usage
```python
import torch
import yaml
from repcodec.RepCodec import RepCodec
# for feature types of HubERT base & data2vec base, please use repcodec_dim768.yaml;
# for feature types of HuBERT large & data2vec large & Whisper medium, please use repcodec_dim1024.yaml;
# for feature types of Whisper large-v2, please use repcodec_dim1280.yaml
config = "repcodec/configs/repcodec_dim768.yaml"
with open(config) as fp:
conf = yaml.load(fp, Loader=yaml.FullLoader)
model = RepCodec(**conf)
model.load_state_dict(torch.load("./hubert_base_l9.pkl", map_location="cpu")["model"]["repcodec"])
model.quantizer.initial()
model.eval()
# input shape: (batch size, hidden dim, sequence length)
random_features = torch.randn(size=(1, 768, 100))
with torch.no_grad():
x = model.encoder(random_features)
z = model.projector(x)
_, idx = model.quantizer.codebook.forward_index(z.transpose(2, 1))
tokens = idx.cpu().data.numpy().tolist()[0]
```
## Training New RepCodec Models
We use a config file to set up all the training configurations, e.g., data, model architecture,
optimizer, scheduler.
We provide an example [here](./train_configs/ex_dim768_mse.yaml).
Please first install required packages following [Installation](#installation)
and prepare the representations following [Representation Preparation](#representation-preparation).
The input data directory is expected to have the following structure
```
/dir/to/representations/
train_set_name/
0_1.npy
0_1.len
valid_set_name/
0_1.npy
0_1.len
test_set_name/
0_1.npy
0_1.len
```
The names of subsets should be the same as the fields in the config file.
Then, you can run training by
```
python train.py \
-c /path/to/config/file \
--tag $tag \
--exp_root exp
```
`tag` is the name of the output folder.
All outputs will be saved to `exp_root/tag/`.
## Acknowledge
Our implementation is based on [facebookresearch/AudioDec](https://github.com/facebookresearch/AudioDec).
We thank them for open-sourcing their code!
## Citation
If you find our work useful, please cite the following article.
```
@misc{huang2023repcodec,
title={RepCodec: A Speech Representation Codec for Speech Tokenization},
author={Zhichao Huang and Chutong Meng and Tom Ko},
year={2023},
eprint={2309.00169},
archivePrefix={arXiv},
primaryClass={eess.AS}
}
```