This repo contains code for training and finetuning Octo generalist robotic policies (GRPs). Octo models are transformer-based diffusion policies, trained on a diverse mix of 800k robot trajectories.
Follow the installation instructions, then load a pretrained Octo model! See examples for guides to zero-shot evaluation and finetuning and
for an inference example.
from octo.model.octo_model import OctoModel
model = OctoModel.load_pretrained("hf:https://rail-berkeley/octo-base")
print(model.get_pretty_spec())
Out of the box, Octo supports multiple RGB camera inputs, can control various robot arms, and can be instructed via language commands or goal images. Octo uses a modular attention structure in its transformer backbone, allowing it to be effectively finetuned to robot setups with new sensory inputs, action spaces, and morphologies, using only a small target domain dataset and accessible compute budgets.
conda create -n octo python=3.10
conda activate octo
pip install -e .
pip install -r requirements.txt
For GPU:
pip install --upgrade "jax[cuda11_pip]==0.4.20" -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html
For TPU
pip install --upgrade "jax[tpu]==0.4.20" -f https://storage.googleapis.com/jax-releases/libtpu_releases.html
See the Jax Github page for more details on installing Jax.
Test the installation by finetuning on the debug dataset:
python scripts/finetune.py --config.pretrained_path=hf:https://rail-berkeley/octo-small --debug
You can find pretrained Octo checkpoints here. At the moment we provide the following model versions:
Model | Inference on 1x NVIDIA 4090 | Size |
---|---|---|
Octo-Base | 13 it/sec | 93M Params |
Octo-Small | 17 it/sec | 27M Params |
We provide simple example scripts that demonstrate how to use and finetune Octo models, as well as how to use our data loader independently. We provide the following examples:
Octo Inference | Minimal example for loading and running a pretrained Octo model |
Octo Finetuning | Minimal example for finetuning a pretrained Octo models on a small dataset with a new observation and action space |
Octo Rollout | Run a rollout of a pretrained Octo policy in a Gym environment |
Octo Robot Eval | Evaluate a pretrained Octo model on a real WidowX robot |
OpenX Dataloader Intro | Walkthrough of the features of our Open X-Embodiment data loader |
OpenX PyTorch Dataloader | Standalone Open X-Embodiment data loader in PyTorch |
To reproduce our Octo pretraining on 800k robot trajectories, run:
python scripts/train.py --config scripts/configs/octo_pretrain_config.py:<size> --name=octo --config.dataset_kwargs.oxe_kwargs.data_dir=... --config.dataset_kwargs.oxe_kwargs.data_mix=oxe_magic_soup ...
To download the pretraining dataset from the Open X-Embodiment Dataset, install the rlds_dataset_mod package and run the prepare_open_x.sh script. The total size of the pre-processed dataset is ~1.2TB.
We run pretraining using a TPUv4-128 pod in 8 hours for the Octo-S model and in 14 hours for Octo-B.
We provide a minimal example for finetuning with a new observation and action space.
We also provide a more advanced finetuning script that allows you to change hyperparameters via a config file and logs finetuning metrics. To run advanced finetuning, use:
python scripts/finetune.py --config.pretrained_path=hf:https://rail-berkeley/octo-small
We offer three finetuning modes depending on the parts of the model that are kept frozen: head_only
, head_mlp_only
, and full
to finetune the full model.
Additionally, one can specify the task type to finetune with: image_conditioned
, language_conditioned
or multimodal
for both.
For example, to finetune the full transformer with image inputs only use:
--config=finetune_config.py:full,image_conditioned
.
Loading and running a trained Octo model is as easy as:
from octo.model import OctoModel
model = OctoModel.load_pretrained("hf:https://rail-berkeley/octo-small")
task = model.create_tasks(texts=["pick up the spoon"])
action = model.sample_action(observation, task, rng=jax.random.PRNGKey(0))
We provide examples for evaluating Octo in a simulated Gym environment as well as on a real WidowX robot.
To evaluate on your own environment, simply wrap it in a Gym interface and follow the instructions in the Eval Env README.
File | Description | |
---|---|---|
Hyperparameters | config.py | Defines all hyperparameters for the training run. |
Pretraining Loop | train.py | Main pretraining script. |
Finetuning Loop | finetune.py | Main finetuning script. |
Datasets | dataset.py | Functions for creating single / interleaved datasets + data augmentation. |
Tokenizers | tokenizers.py | Tokenizers that encode image / text inputs into tokens. |
Octo Model | octo_model.py | Main entry point for interacting with Octo models: loading, saving, and inference. |
Model Architecture | octo_module.py | Combines token sequencing, transformer backbone and readout heads. |
Visualization | visualization_lib.py | Utilities for offline qualitative & quantitative eval. |
@misc{octo_2023,
title={Octo: An Open-Source Generalist Robot Policy},
author = {{Octo Model Team} and Dibya Ghosh and Homer Walke and Karl Pertsch and Kevin Black and Oier Mees and Sudeep Dasari and Joey Hejna and Charles Xu and Jianlan Luo and Tobias Kreiman and {You Liang} Tan and Dorsa Sadigh and Chelsea Finn and Sergey Levine},
howpublished = {\url{https://octo-models.github.io}},
year = {2023},
}