Fine-tune the Voxtral speech model for automatic speech recognition (ASR) using Hugging Face transformers and datasets. The recommended way to run training is Hugging Face Jobs: push your dataset to the Hub, then launch training on HF infrastructure (default a100-large GPU) with one script—no local GPU required.
The scripts/launch_hf_job.py script submits your training run to Hugging Face Jobs.… See the full description on the dataset page:
https://huggingface.co/datasets/shakods/synthetic-data.