tduggan93/pi0-hanoi-pg-30k

PI0 Hanoi Policy Gradient Checkpoint (30k steps)

This is a checkpoint for the PI0 (Physical Intelligence 0) model trained on the Hanoi task using a subtask-based approach with policy gradient methods.

Model Details

• Task: Hanoi Tower puzzle (subtask decomposition)
• Training Steps: 30,000
• Model Type: Policy gradient with subtask learning
• Framework: JAX/Flax
• Dataset: hanoi_300_lerobot
• Architecture: Vision-Language-Action model with subtask masking

Key Features

• Subtask Learning: Decomposes Hanoi puzzle into manageable subtasks
• No Masking: Trained without subtask masking for better generalization
• Policy Gradient: Uses policy gradient optimization
• End-to-End: Learns from visual observations to actions

Checkpoint Structure

• params/: Model parameters
• train_state/: Training state
• assets/: Additional assets including normalization statistics
• _CHECKPOINT_METADATA: Checkpoint metadata

Usage

This checkpoint can be loaded using the appropriate JAX/Flax model loading utilities in your training pipeline.

Training Configuration

• Dataset: hanoi_300_lerobot
• Training approach: Subtask-based policy gradient
• Subtask masking: Disabled (no_masking)
• Total training steps: 30,000
• Model: PI0 with subtask decomposition

Model Card

This model is trained for the Hanoi Tower puzzle task using a subtask-based policy gradient approach. The model decomposes the complex Hanoi puzzle into simpler subtasks and learns to solve them sequentially.

Performance

• Training Steps: 30,000
• Architecture: Vision-Language-Action model with subtask learning
• Approach: Policy gradient without subtask masking

Advantages

• Subtask Decomposition: Breaks down complex tasks into manageable parts
• Better Generalization: No masking allows for more flexible learning
• Policy Gradient: Direct optimization of policy performance
• End-to-End: Learns complete task from visual input

Limitations

• Trained specifically on Hanoi Tower puzzle with subtask structure
• Performance may vary on different manipulation tasks
• Requires appropriate normalization statistics for inference
• Subtask decomposition may not generalize to all manipulation tasks

Technical Details

• Subtask Approach: Decomposes Hanoi puzzle into logical subtasks
• No Masking: Allows model to learn across subtask boundaries
• Policy Gradient: Direct policy optimization without value function
• Vision Input: Processes visual observations from robot cameras
• Action Output: Generates robot manipulation actions