Files
MUSE-Books_iclm-7b_npo_beta…/README.md

22 lines
434 B
Markdown
Raw Normal View History

---
library_name: transformers
base_model: muse-bench/MUSE-books_target
tags:
- MUSE
- unlearning
- books
---
# Jeesup/MUSE-Books_Llama-2-7b_npo_beta0p1_lr5e-6_lam10_ep1
MUSE unlearning run uploaded by `run_muse_settings_to_hf.py`.
- Corpus: books
- Algorithm: npo (`npo_gdr`)
- Learning rate: 5e-06
- Retain loss coefficient lambda: 10.0
- Epochs: 1
- NPO beta: 0.1
See `muse_run_metadata.json` for checkpoint selection details.