GitHub - sakethyalamanchili/Implementation_of_QLoRA_with_BERT_on_AGNews: Fine-tuning doesn't require a server cluster! This project uses QLoRA to optimize BERT for news classification. By combining 4-bit quantization and LoRA, I slashed VRAM usage by 75% and trained only 2% of parameters. The result? A pro-level classifier in a tiny 20MB adapter file—high performance, minimal hardware.

Branches Tags

Name		Name	Last commit message	Last commit date
Latest commit History 3 Commits
implementation-of-qlora-with-bert-on-agnews.ipynb		implementation-of-qlora-with-bert-on-agnews.ipynb

About

Fine-tuning doesn't require a server cluster! This project uses QLoRA to optimize BERT for news classification. By combining 4-bit quantization and LoRA, I slashed VRAM usage by 75% and trained only 2% of parameters. The result? A pro-level classifier in a tiny 20MB adapter file—high performance, minimal hardware.

www.kaggle.com/code/sakethyalamanchili/implementation-of-qlora-with-bert-on-agnews

pytorch transformer accelerate lora bert agnews qlora bitsandbytes