Key Info

Hugging Face now streams live training runs on its platform, with the first showcased run training LoRA adapters via reinforcement learning on Liquid AI's LFM2.5/LFM3 and thewebAI's Twil-9B model.

Highlights

  • Live training runs can be followed directly on Hugging Face
  • The initial run uses RL to train LoRA adapters
  • Models involved: LFM2.5 and LFM3 from Liquid AI, plus Twil-9B from thewebAI
  • Goal is more efficient and accurate performance on agentic tasks