Tagged: trl

7 articles

Latest

DPO Cuts RLHF Down to a Binary Cross-Entropy Loss for Llama 2

Stable Diffusion gets RLHF-style fine-tuning with TRL's new DDPOTrainer

RLHF on a Budget: Qwen2.5-0.5B Learns Preferences in 30 Steps

Unsloth claims 2.7x faster LLM fine-tuning with zero accuracy loss

Async RL's dirty secret: 98% of weights don't change between steps, and TRL now exploits that

DeepSpeed ZeRO-3 clashing with Liger GRPO loss? Shape mismatch halts Qwen2.5 finetuning

Hugging Face TRL now trains vision models with DPO on 83K-row datasets