Flower Calvin D

🚀 FlowerVLA - Vision-Language-Action Flow Model for CALVIN D

A pretrained FlowerVLA model for robotic manipulation, trained on the CALVIN D dataset. Flower is an efficient Vision-Language-Action Flow policy for robot learning, containing only 1B parameters.

🚀 Quick Start

Check out our full model implementation on Github todo and follow the instructions in the readme to test the model on one of the environments.

obs = {
    "rgb_obs": {
        "rgb_static": static_image,
        "rgb_gripper": gripper_image
    }
}
goal = {"lang_text": "pick up the blue cube"}
action = model.step(obs, goal)

✨ Features

FlowerVLA is a novel architecture that:

• Uses half of Florence-2 for multi-modal vision-language encoding
• Employs an novel transformer-based flow matching architecture
• Provides an efficient, versatile VLA policy with only ~1B parameters

📦 Installation

No installation steps provided in the original document, so this section is skipped.

💻 Usage Examples

Basic Usage

obs = {
    "rgb_obs": {
        "rgb_static": static_image,
        "rgb_gripper": gripper_image
    }
}
goal = {"lang_text": "pick up the blue cube"}
action = model.step(obs, goal)

Advanced Usage

No advanced usage code provided in the original document, so this part is skipped.

📚 Documentation

Model Performance

This checkpoint contains weights for the CALVIN D challenge and currently ranks 1 with the following results:

Input/Output Specifications

Inputs

• RGB Static Camera: (B, T, 3, H, W) tensor
• RGB Gripper Camera: (B, T, 3, H, W) tensor
• Language Instructions: Text strings

Outputs

• Action Space: (B, T, 7) tensor representing delta EEF actions

Training Details

Configuration

• Optimizer: AdamW
• Learning Rate: 2e-5
• Weight Decay: 0.05

🔧 Technical Details

No specific technical details (more than 50 words) provided in the original document, so this section is skipped.

📄 License

This model is released under the MIT license.

@inproceedings{ reuss2025flower, # Add citation when available }