payelb's picture
Upload PPO-aligned TinyLlama-1.1B model using MARS reward model on HHRLHF
a75499e verified

payelb/HHRLHF_TinyLlama-1.1B_aligned_with_MARS_RM

Base model: TinyLlama/TinyLlama-1.1B-Chat-v1.0

Alignment dataset: Anthropic/hh-rlhf

Reward model: payelb/HHRLHF_roberta-base_1k_fixed_MARS

Method: PPO alignment with LoRA adapters.

Notes:

  • Reward normalization and clipping enabled
  • KL control enabled
  • pad_token_id/eos_token_id explicitly set