GR00T Policy Fine-tuning

Unleash advanced robotic control with GR00T Policy Fine-tuning. Ready to experience the power of AI? Start your journey here!

Platform: Replicate

Robotics TrainingPolicy Fine-tuningGR00T-N1 FrameworkLeRobot Dataset

263 runs

A100 (80GB)

roughly $10 per training run

License Check Required

🚀Function Overview

A pipeline for fine-tuning NVIDIA's GR00T-N1 foundation model on LeRobot-formatted datasets to generate robotic control policies.

Key Features

Fine-tunes robotic control policies using GR00T-N1
Supports LeRobot v2.0+ datasets with 1-2 cameras and 6-DoF robots
Automatically pushes trained models to Hugging Face Hub
Integrates with Weights & Biases for training tracking
Configurable training parameters (batch size, epochs, learning rate)

Use Cases

•Creating customized robotic control models
•Research in reinforcement learning for robotics
•Adapting foundation models for specific robot configurations
•Training automation pipelines for AI-driven robotics

⚙️Input Parameters

dataset_repo_id

string

Hugging Face dataset ID to train on, LeRobot format > v2.0 expected, i.e. 'LegrandFrederic/dual-setup'

hugging_face_token

string

Hugging Face API token, needed to download your dataset and upload your model. Please use a fine grained token with Repository write access.

wandb_api_key

string

Weights & Biases API key (optional, to track the online training), find yours here: https://wandb.ai/authorize

hf_model_name

string

Hugging Face model name to upload the trained model to (optional). If not provided, a random name will be generated.

batch_size

integer

Batch size for training

number_of_epochs_to_train_for

integer

Number of epochs to train for

learning_rate

number

Learning rate for training

modal_token_id

string

Modal token ID, if you don't know what this is, leave it empty

modal_token_secret

string

Modal token secret, if you don't know what this is, leave it empty

💡Usage Examples

Example 1

Input Parameters

{
  "batch_size": 32,
  "learning_rate": 0.0002,
  "dataset_repo_id": "00ri/so100_battery",
  "hugging_face_token": "[REDACTED]",
  "hugging_face_model_name": "00ri/so100_battery-ovrr0igj39",
  "number_of_epochs_to_train_for": 10
}

Output Results

https://replicate.delivery/yhqm/EeA0vq3JepvcOE1s6cKKcFk0TplWSX8cybeTiuMhmO5fAt1RB/README.md

Quick Actions

Use NowView Documentation

Technical Specifications

Hardware Type: A100 (80GB)
Run Count: 263
Commercial Use: Unknown/Restricted
Pricing: roughly $10 per training run
Platform: Replicate

Related Keywords

Robotic Control PoliciesLeRobot DatasetsHugging Face Hub IntegrationWeights & Biases TrackingCustomized Robotic Control ModelsAI-Driven RoboticsReinforcement Learning Research

Related Models

TinyClick GUI Automation

TinyClick: Single-Turn Agent for Empowering GUI Automation