lerobot/pi0_libero_base

Full fine-tune vs lerobot/pi0_base. Fine-tune: 100% of the weights changed, across 27 of 27 layers.

Tensors changed: other 8/8, MLP 220/220, lm_head 2/2, Norms 184/184, Attention 360/360, Embeddings 3/3

1 derivatives checked

On the Hub · Downloads on Model Pulse