Skip to content

About

No description, website, or topics provided.

Resources

Code of conduct

Contributing

Security policy

Stars

3 stars

Watchers

3 watching

Forks

Luciole-Audio-Training

Training and export scripts for Luciole Audio: speech/audio-augmented LLMs (SALM) built on the Luciole models (1B, 8B, 23B) with NVIDIA NeMo speechlm2.

Repository layout

Folder Content
data/ Dataset conversion to NeMo/Lhotse manifests, data generation, templating, TTS
training/ SALM training (salm_train.py), Hydra configs (conf/), Slurm launcher
export/ Export of NeMo checkpoints to Hugging Face format (to_hf.py)
utils/ Shared helpers (audio, inference, metrics)
assets/ Model/dataset cards, plotting and upload scripts
demo/ WebRTC demo app
install/ Environment installation scripts (Jean-Zay)

Installation

Data preparation scripts (data/) import luciole_audio.*, so the repo must be installed:

pip install -e ".[data]"   # or [generation], [processing], [templating], [all-data], ...

The tts extra conflicts with the others (different transformers pin) and must be installed in a separate environment.

Usage

Training (submits training/run_train.slurm):

python training/slurm_launcher.py --config run/luciole-8b --gpus 4 --time 20:00:00

Hugging Face export:

sbatch export/run_hf_export.slurm

Evaluation

Models are evaluated with our fork of AudioBench (results on the leaderboard).

License

AGPL v3, see LICENSE.

See NOTICE for copyright and attribution, including the ownership and licenses of the Luciole Audio models and datasets.

About

No description, website, or topics provided.

Resources

Code of conduct

Contributing

Security policy

Stars

3 stars

Watchers

3 watching

Forks

Releases

Packages

Contributors

Languages