Vaibhavs10/fast-whisper-finetuning

This is stars and forks stats for /Vaibhavs10/fast-whisper-finetuning repository. As of 12 May, 2024 this repository has 280 stars and 19 forks.

Faster Whisper Finetuning with LoRA powered by 🤗 PEFT TL;DR - A one size fits all walkthrough, to fine-tune Whisper (large) 5x faster on a consumer GPU with less than 8GB GPU VRAM, all with comparable performance to full-finetuning. ⚡️ Not convinced? Here are some benchmarks we ran on a free Google Colab T4 GPU! 👇 Training type Trainable params Memory allocation Max. batch size LoRA <1% 8GB 24 adaLoRA <0.9% 7.9GB 24 Full-fine-tuning 100% OOM on T4 OOM on T4 Table of Contents Why Parameter...