OneReason-0.8B Online GRPO LoRA - Epoch 2
This repository contains one LoRA adapter checkpoint for the Kuaishou LLM4Rec competition.
Training
- Start: OneReason-0.8B competition base with the approved SFT adapter
- Method: G=8 live constrained sampling, forced-GT safeguard, two epochs
- Training groups: 6,378 groups x 2 epochs
- Completed epoch: 2
- LoRA rank / alpha / dropout: 32 / 32 / 0.0
- Target modules: q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj
Files
adapter_model.safetensors: LoRA weightsadapter_config.json: the original training checkpoint config, preserved byte-for-byte
Evaluation status
A local fixed beam-16 constrained probe was completed; no official competition score is claimed.
