EN

Octo Model Team

Octo-Small 1.5

Octo-Small 1.5

Checkpoint

Transformer diffusion policy pretrained on 800K Open X-Embodiment episodes.

Specifications

Architecture components: Diffusion action decoder
Backbone: Transformer-based diffusion policy
Frameworks: JAX / Octo
Input modalities: Goal image
Input modalities: Language instruction
Input modalities: RGB camera images
Input modalities: Robot state / proprioception
Languages: English
Model type: Vision-language-action model
Output action space: Flexible continuous robot actions
Parameter count: 0.027B parameters
Pipeline task: Robotics
Training data scale: 800K robot episodes
Training data sources: Open X-Embodiment

Lineage

Unknown

Architecture & I/O

trajectory · continuous

Unknown

Flexible continuous robot actions

Source · rail-berkeley

Default checkpoint

rail-berkeley/octo-small-1.5

openMIT License

Other releases

1

published checkpoints

Supported robots

No verified robot compatibility recorded.

Benchmark results

No published benchmark results.

Deployment

JAX / Octo · NVIDIA GPU

bfloat16 · unknown

Source · rail-berkeley

Resources

paper · Octo paper project page · Octo project page project page · Octo-Small 1.5 training data source repository · Octo repository weights · rail-berkeley/octo-small-1.5 model card

Submit info