hppRC/llm-lora-classification

LLMとLoRAを用いたテキスト分類

34
/ 100
Emerging

Implements LoRA-based fine-tuning for Japanese LLMs on text classification by leveraging the final token embedding (rather than `[CLS]`) as the classification representation, accounting for unidirectional model architecture. Integrates with Hugging Face's `transformers` and `PEFT` libraries, evaluated on 7 Japanese LLM variants (rinna, CyberAgent) with systematic hyperparameter exploration across learning rates and input templates. Achieves 97.75% F1 on 9-class livedoor news classification, outperforming comparable BERT baselines despite using comparable model scales.

No commits in the last 6 months.

Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 9 / 25
Maturity 16 / 25
Community 9 / 25

How are scores calculated?

Stars

98

Forks

6

Language

Python

License

MIT

Last pushed

Jul 22, 2023

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/llm-tools/hppRC/llm-lora-classification"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.