Fine-tune Whisper Model
Welcome to our comprehensive guide for fine-tuning Whisper models! This guide provides a detailed walkthrough to help you launch, track, and understand the billing process associated with fine-tuning Whisper on the MonsterAPI Platform.
Demo Colab Notebooks with Python Client
Follow this guide to initiate a Whisper fine-tuning job. It covers everything from opening the fine-tuning portal, creating a new job, selecting a model and dataset, to submitting the job.
Supported Models for Finetuning
Model | Model | Model |
---|---|---|
OpenAI/whisper-large-v3 | OpenAI/whisper-large | OpenAI/whisper-base |
OpenAI/whisper-tiny | OpenAI/whisper-small.en | OpenAI/whisper-tiny.en |
OpenAI/whisper-medium | OpenAI/whisper-small | OpenAI/whisper-medium.en |
OpenAI/whisper-large-v2 | OpenAI/whisper-base.en | distil-whisper/distil-small.en |
distil-whisper/distil-medium.en | distil-whisper/distil-large-v2 |
Choose an appropriate dataset from Hugging Face. Ensure that audio datasets include a column for audio files and a corresponding text column. If the dataset is private, provide your Hugging Face "Read Key" for access. For uploading fine-tuning content, include your Hugging Face "Write Key."
After launching your fine-tuning job, monitor its progress using this guide. Learn about job stages, view logs, track metrics with Weights & Biases (if credentials are provided), and download your fine-tuned model upon completion.
Understand the costs associated with fine-tuning jobs. This guide covers job billing, per-minute costs, handling credit depletion, and maintaining an active payment method and subscription plan.
For further questions or assistance, contact us at MonsterAPI Support.
Updated 4 months ago