HelloCephalopod/my_smolvla_2

SmolVLA Fine-tuned Model

This model is a fine-tuned version of HuggingFaceTB/SmolVLA trained using the LeRobot library.

Model Details

• Base Model: HuggingFaceTB/SmolVLA
• Training Steps: 20,000
• Batch Size: 10
• Checkpoints Available: 4

Available Checkpoints

This repository contains multiple checkpoints from the training process:

• checkpoint-020000/ - Step 020000
• checkpoint-040000/ - Step 040000
• checkpoint-060000/ - Step 060000
• checkpoint-last/ - Step last

Usage

from lerobot import LeRobotPolicy

# Load the final checkpoint
model = LeRobotPolicy.from_pretrained("HelloCephalopod/my_smolvla_2/checkpoint-last")

# Or load a specific checkpoint
model = LeRobotPolicy.from_pretrained("HelloCephalopod/my_smolvla_2/checkpoint-10000")

Training Details

• Training Data: Custom dataset from Hugging Face Hub
• Training Steps: 60,000
• Batch Size: 10
• Save Steps: 20,000
• Hardware: Google Colab GPU

Checkpoint Selection

• checkpoint-40000: Mid-training checkpoint
• checkpoint-60000: Final checkpoint
• checkpoint-last: Latest checkpoint (recommended)

Citation

If you use this model, please cite the original SmolVLA paper and the LeRobot library.