Model Trainer

The Model Trainer skill enables training and fine-tuning language models using TRL (Transformer Reinforcement Learning) on Hugging Face Jobs managed infrastructure. Supports multiple training methods including SFT for instruction tuning, DPO for preference optimization, GRPO for online RL, and reward modeling for RLHF. Includes complete workflow management with dataset validation, hardware selection, cost estimation, Trackio monitoring, Hub integration, and GGUF conversion for local deployment.