YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

finetuned_with_lift_pi05

Private backup of a Pi0.5 Lift behavior-cloning fine-tune checkpoint.

Base

  • Base model path during training: /data/ubuntu/models/pi05/pi05_base
  • Policy type: LeRobot PI05Policy / Pi0.5-style model

Dataset

  • Source: MIML-VLA Lift HDF5
  • Converted dataset path: /data/ubuntu/miml_vla/suturing_pi05/data/pi05_lift_32d
  • Total samples: 2368
  • Train samples: 2132
  • Validation samples: 236
  • Original state: 24D
  • Padded state: 32D
  • Original action: 7D
  • Padded action: 32D
  • Action chunk size: 50

Training

  • Steps: 100
  • Trainable mode: last_layers
  • Trainable parameters: about 263M / 4.14B
  • Batch size: 1
  • Learning rate: 1e-6
  • Final train loss: 0.025313330814242363
  • Internal validation loss: 0.057169350981712344

Notes

This is the behavior-cloning fine-tuned checkpoint before online RLT. It is intended as the starting checkpoint for rlt_finetuned_pi05.

Downloads last month
9
Safetensors
Model size
4B params
Tensor type
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support