Cogact Large

🚀 CogACT-Large

CogACT is a novel and advanced Vision-Language-Action (VLA) architecture developed from Vision-Language Model (VLM). Different from previous approaches that simply repurpose VLM for action prediction through basic action quantization, we present a componentized VLA architecture featuring a specialized action module based on VLM output. CogACT-Large utilizes a DiT-L model as its action module.

All our code and pretrained model weights are licensed under the MIT license.

For more details, please refer to our project page and paper.

🚀 Quick Start

CogACT takes a language instruction and a single - view RGB image as input and predicts the next 16 normalized robot actions (comprising the 7 - DoF end - effector deltas in the form of x, y, z, roll, pitch, yaw, gripper). These actions should be unnormalized and integrated by our Adaptive Action Ensemble (Optional). Unnormalization and ensemble rely on the dataset statistics.

CogACT models can be used zero - shot to control robots for setups seen in the [Open - X](https://robotics - transformer - x.github.io/) pretraining mixture. They can also be fine - tuned for new tasks and robot setups with an extremely small number of demonstrations. See our repository for more information.

✨ Features

• Advanced Architecture: A componentized VLA architecture with a specialized action module.
• Versatile Usage: Can be used for zero - shot robot control and fine - tuned for new tasks.

📦 Installation

Please clone and install dependencies in our repo. Install minimal dependencies (torch, transformers, timm, tokenizers, ...).

💻 Usage Examples

Basic Usage

# Please clone and install dependencies in our repo
# Install minimal dependencies (`torch`, `transformers`, `timm`, `tokenizers`, ...)

from PIL import Image
from vla import load_vla
import torch

model = load_vla(
      'CogACT/CogACT-Large',
      load_for_training=False,
      action_model_type='DiT-L',
      future_action_window_size=15,
    )                                 
# about 30G Memory in fp32; 

# (Optional) use "model.vlm = model.vlm.to(torch.bfloat16)" to load vlm in bf16

model.to('cuda:0').eval()

image: Image.Image = <input_your_image>
prompt = "move sponge near apple"           # input your prompt

# Predict Action (7-DoF; un-normalize for RT-1 google robot data, i.e. fractal20220817_data)
actions, _ = model.predict_action(
          image,
          prompt,
          unnorm_key='fractal20220817_data', # input your unnorm_key of dataset
          cfg_scale = 1.5,                   # cfg from 1.5 to 7 also performs well
          use_ddim = True,                   # use DDIM sampling
          num_ddim_steps = 10,               # number of steps for DDIM sampling
        )

# results in 7-DoF actions of 16 steps with shape [16, 7]

📚 Documentation

Model Summary

📄 License

All our code and pretrained model weights are licensed under the MIT license.

📄 Citation

@article{li2024cogact,
  title={CogACT: A Foundational Vision-Language-Action Model for Synergizing Cognition and Action in Robotic Manipulation},
  author={Li, Qixiu and Liang, Yaobo and Wang, Zeyu and Luo, Lin and Chen, Xi and Liao, Mozheng and Wei, Fangyun and Deng, Yu and Xu, Sicheng and Zhang, Yizhong and others},
  journal={arXiv preprint arXiv:2411.19650},
  year={2024}
}