Hugging Face Model Trainer

What This Skill Does

Train and fine-tune language models using TRL on Hugging Face Jobs — SFT, DPO, GRPO, reward modeling, and GGUF conversion.

Capabilities

• Fine-Tuning Methods — SFT, DPO, GRPO, and reward modeling
• Cloud GPU Training — Train on HF Jobs with proper hardware selection
• GGUF Conversion — Convert models for local deployment
• Monitoring — Track training progress with Trackio

How It Works

1. Select a base model and training method
2. Prepare and validate your training dataset
3. Submit the training job to HF Jobs with selected hardware
4. Monitor progress and download the fine-tuned model

Example Usage

"Fine-tune Llama with SFT on my instruction dataset"
"Run DPO training for preference alignment"
"Convert my model to GGUF for local inference"