PROMPT="""You will be given six descriptive keywords related to an audio sample of a person's speech. These keywords include:
1. The gender (e.g., male, female)
2. The level of reverberation (e.g., very roomy sounding, quite roomy sounding, slightly roomy sounding, moderate reverberation, slightly confined sounding, quite confined sounding, very confined sounding)
3. The amount of noise the sample (e.g., very noisy, quite noisy, slightly noisy, moderate ambient sound, slightly clear, quite clear, very clear)
4. The tone of the speaker's voice (e.g., very monotone, quite monotone, slightly monotone, moderate intonation, slightly expressive, quite expressive, very expressive)
5. The pace of the speaker's delivery (e.g., very slowly, quite slowly, slightly slowly, moderate speed, slightly fast, quite fast, very fast)
6. The pitch of the speaker's voice (e.g., very low pitch, quite low pitch, slightly low pitch, moderate pitch, slightly high pitch, quite high pitch, very high pitch)
Your task is to create a text description using these keywords that accurately describes the speech sample while ensuring the description remains grammatically correct and easy to understand. You should rearrange the keyword order as necessary, and substitute synonymous terms where appropriate. If the amount of noise is 'very noisy' and the level of reverberation is 'very roomy sounding', include terms like 'very bad recording' in the description. Likewise, if the amount of noise is 'very clear' and the level of reverberation is 'very confined sounding', include terms like 'very good recording' in the description. Otherwise, do not add extra details beyond what has been provided, and only return the generated description.
For example, given the following keywords: 'female', 'slightly roomy sounding', 'slightly noisy', 'very expressive', 'slightly low pitch', 'very slowly', a valid description would be: 'a woman with a deep voice speaks slowly but has an animated delivery in an echoey room with some background noise'.
For the keywords: '[gender]', '[reverberation]', '[noise]', '[speech_monotony]', '[pitch]', '[speaking_rate]', the corresponding description is:"
**TL;DR:** After having followed the [installation steps](#requirements), you can reproduce the Parler-TTS v0.1 training recipe with the following command line:
This sub-folder contains all the information to train or fine-tune your own Parler-TTS model. It consists of:
-[1. An introduction to the Parler-TTS architecture](#a-architecture)
-[2. First steps to get started](#b-getting-started)
-[3. Training guide](#c-training)
## 1. Architecture
At the moment, Parler-TTS architecture is a carbon copy of the [MusicGen architecture](https://huggingface.co/docs/transformers/v4.39.3/en/model_doc/musicgen#model-structure) and can be decomposed into three distinct stages:
1. Text encoder: maps the text descriptions to a sequence of hidden-state representations. Parler-TTS uses a frozen text encoder initialised entirely from Flan-T5
2. Parler-TTS decoder: a language model (LM) that auto-regressively generates audio tokens (or codes) conditional on the encoder hidden-state representations
3. Audio codec: used to recover the audio waveform from the audio tokens predicted by the decoder. We use the [DAC model](https://github.com/descriptinc/descript-audio-codec) from Descript, although other codec models, such as [EnCodec](https://huggingface.co/facebook/encodec_48khz), can also be used
Parler-TTS however introduces some small tweaks:
- The text **description** is passed through the text encoder and used in the cross-attention layers of the decoder.
- The text **prompt** is simply passed through an embedding layer and concatenated to the decoder input hidden states.
- The audio encoder used is [**DAC**](https://descript.notion.site/Descript-Audio-Codec-11389fce0ce2419891d6591a68f814d5) instead of [Encodec](https://github.com/facebookresearch/encodec), as it exhibits better quality.
## 2. Getting started
To get started, you need to follow a few steps:
1. Install the requirements.
2. Find or initialize the model you'll train on.
3. Find and/or annotate the dataset you'll train your model on.
### Requirements
The Parler-TTS code is written in [PyTorch](https://pytorch.org) and [Accelerate](https://huggingface.co/docs/accelerate/index). It uses some additional requirements, like [wandb](https://wandb.ai/), especially for logging and evaluation.
To install the package for training, you need to clone the repository from source...
Optionally, you can create a wandb account and login to it by following [this guide](https://docs.wandb.ai/quickstart). [`wandb`](https://docs.wandb.ai/) allows for better tracking of the experiments metrics and losses.
You also have the option to configure Accelerate by running the following command. Note that you should set the number of GPUs you wish to use for training, and also the data type (dtype) to your preferred dtype for training/inference (e.g. `bfloat16` on A100 GPUs, `float16` on V100 GPUs, etc.):
```bash
accelerate config
```
Lastly, you can link you Hugging Face account so that you can push model repositories on the Hub. This will allow you to save your trained models on the Hub so that you can share them with the community. Run the command:
```bash
git config --global credential.helper store
huggingface-cli login
```
And then enter an authentication token from https://huggingface.co/settings/tokens. Create a new token if you do not have one already. You should make sure that this token has "write" privileges.
### Initialize a model from scratch or use a pre-trained one.
Depending on your compute resources and your dataset, you need to choose between fine-tuning a pre-trained model and training a new model from scratch.
In that sense, we released a 300M checkpoint trained on 10.5K hours of annotated data under the repository id: [`parler-tts/parler_tts_300M_v0.1`](https://huggingface.co/parler-tts/parler_tts_300M_v0.1), that you can fine-tune for your own use-case.
You can also train you own model from scratch. You can find [here](/helpers/model_init_scripts/) examples on how to initialize a model from scratch. For example, you can initialize a dummy model with:
To train your own Parler-TTS, you need datasets with 3 main features:
- speech data
- text transcription of the speech data
- conditionning text description - that you can create using [Data-Speech](https://github.com/huggingface/dataspeech), a library that allows you to annotate the speaker and utterance characteristics with natural language description.
Note that we made the choice to use description of the main speech characteristics (speaker pitch, speaking rate, level of noise, etc.) but that you are free to use any handmade or generated text description that makes sense.
To train Parler-TTS v0.1, we used:
* The full [LibriTTS-R dataset](https://huggingface.co/datasets/blabble-io/libritts_r), a 1K hours high-quality speech dataset.
* A [10K hours subset](https://huggingface.co/datasets/parler-tts/mls_eng_10k) of [Multilingual LibriSpeech](https://huggingface.co/datasets/facebook/multilingual_librispeech).
Both datasets have been annotated using the [Data-Speech](https://github.com/huggingface/dataspeech) recipe, respectively [here](https://huggingface.co/datasets/parler-tts/libritts_r_tags_tagged_10k_generated) and [here](https://huggingface.co/datasets/parler-tts/mls-eng-10k-tags_tagged_10k_generated).
## 3. Training
The script [`run_parler_tts_training.py`](/training/run_parler_tts_training.py) is an end-to-end script that:
1. load dataset(s) and merge them to the annotation dataset(s) if necessary
In particular, note how multiple training datasets, metadataset, configurations and splits can be loaded by separating the dataset arguments by + symbols:
Additionally, you can also write a JSON config file. Here, [starting_point_0.01.json](helpers/training_configs/starting_point_0.01.json) contains the exact same hyper-parameters than above and can be launched like that:
Training logs will be reported to wandb, provided that you passed `--report_to "wandb"` to the arguments. An example of what a training log from the above training looks like can be found [here](https://wandb.ai/ylacombe/parler-tts-300M-punctuated/runs/q6h7hspc?nw=nwuserylacombe).
> [!TIP]
> Starting training a new model from scratch can easily be overwhelming, so here's what training looked like for v0.1: [logs](https://api.wandb.ai/links/ylacombe/ea449l81)
Scaling to multiple GPUs using [distributed data parallelism (DDP)](https://pytorch.org/tutorials/beginner/ddp_series_theory.html) is trivial: simply run `accelerate config` and select the multi-GPU option, specifying the IDs of the GPUs you wish to use. The above script can then be run using DDP with no code changes. In our case, we used a node of 8 H100 80GB to train Parler-TTS v0.1 for around 4 days.
There are a few other noteworthy arguments:
1.`train_metadata_dataset_name` and `eval_metadata_dataset_name` specify, if necessary, the names of the dataset(s) that contain(s) the conditionning text descriptions. For example, this [dataset resulting from the Data-Speech annotation process](https://huggingface.co/datasets/parler-tts/libritts_r_tags_tagged_10k_generated) is saved without the audio column, as it's costly to write and push audio data, so it needs to be concatenated back to the original LibriTTS-R dataset.
2. As noted above, the script pre-computes audio tokens as computing audio codes is costly and only needs to be done once, since we're freezing the audio encoder. `audio_encoder_per_device_batch_size` is used to precise the per devie batch size for this pre-processing step.
3. Additionnally, when scaling up the training data and iterating on the hyper-parameters or the model architecture, we might want to avoid recomputing the audio tokens at each training run. That's why we introduced two additional parameters, `save_to_disk` and `temporary_save_to_disk` that serves as temporary buffers to save intermediary datasets. Note that processed data is made of text and audio tokens which are much more memory efficient, so the additional required space is negligible.
4.`predict_with_generate` and `add_audio_samples_to_wandb` are required to store generated audios and to compute WER and CLAP similarity.
5.`freeze_text_encoder`: which allows to freeze the text encoder, to save compute resources.
And finally, two additional comments:
1.`lr_scheduler_stype`: defines the learning rate schedule, one of `constant_with_warmup` or `cosine`. When experimenting with a training set-up or training for very few epochs, using `constant_with_warmup` is typically beneficial, since the learning rate remains high over the short training run. When performing longer training runs, using a `cosine` schedule shoud give better results.
2.`dtype`: data type (dtype) in which the model computation should be performed. Note that this only controls the dtype of the computations (forward and backward pass), and not the dtype of the parameters or optimiser states.
> [!TIP]
> Fine-tuning is as easy as modifying `model_name_or_path` to a pre-trained model.
> For example: `--model_name_or_path parler-tts/parler_tts_300M_v0.1`.
