Files
Mistral-7B-v0.1-DPO/README.md

8 lines
349 B
Markdown
Raw Normal View History

---
license: apache-2.0
---
Mistral-7b-v0.1-DPO is a finetuned adapter from the original Mistral-7b model. In this adaptor, I am finetuning the LM head in addition to the regular modules that are normally finetuned. Below is the list of the finetuned modules:
'k_proj', 'gate_proj', 'v_proj', 'up_proj', 'q_proj', 'o_proj', 'down_proj', 'lm_head'