Auto_ViML builds multiple machine learning models with a single line of code.
Open source
24 alternativesOcto
Octo provides a transformer-based policy for robots, with code to train and fine-tune it.
About Octo
Octo is an open-source repository with code to train and finetune transformer-based diffusion policies for robots. The policies were trained on 800k robot trajectories and support multiple RGB camera inputs, various robot arms, and instruction via language commands or goal images.
The repository provides pretrained Octo-Base (93M parameters) and Octo-Small (27M parameters) checkpoints, example scripts for inference, finetuning, Gym rollouts, and evaluation on a WidowX robot, and scripts to reproduce pretraining. Reproducing pretraining requires about 1.2TB of preprocessed Open X-Embodiment data and TPU resources.
Key features
- Load pretrained Octo-Base or Octo-Small checkpoints via Python
- Control robot arms using language commands or goal images
- Accept multiple RGB camera inputs for observation
- Finetune with head_only, head_mlp_only, or full modes
- Finetune for image, language, or multimodal task types
- Run rollout in Gym environment or evaluate on WidowX robot
Best for: Researchers and developers who train or adapt robot control policies for different arms and inputs.
Good for
- Includes code for training, finetuning, inference, and data loading
- Provides two pretrained checkpoints with reported inference rates
- Supports finetuning to new sensory inputs, action spaces, and morphologies
Watch out for
- Reproducing pretraining needs about 1.2TB of preprocessed data
- Pretraining used a TPUv4-128 pod, with 8-14 hours reported
- Installation requires Python 3.10 and a separate JAX install for GPU or TPU
Octo at a glance
- Website
- github.com/octo-models/octo
- Type
- Open-source project
- Pricing
- Not confirmed. Check the site
- Open source
- Yes · Source code
- Language
- English
- Last checked
- Oct 6, 2026 · Online when checked
- Listed
- Oct 5, 2026
Octo alternatives
Tools and sites you could use instead, ranked by how closely they match.
m2cgen converts machine learning models into native code for multiple programming languages without runtime dependencies.
Open source
24 alternatives
PyTorch Lightning extension that accelerates and enhances foundation model experimentation with flexible fine-tuning schedules.
Open source
24 alternatives
Hugging Face Accelerate launches and trains PyTorch models across devices and distributed configurations.
Open source
24 alternatives
Use Learnergy, a Python library for training and working with energy-based machine learning models.
Open source
24 alternatives
LightlyTrain supports pretraining, fine-tuning, and distillation for computer vision models.
Open source
24 alternativesPopular in ML Training & Optimization
Well-known listings from the same topic.
Edwardedwardlib.orgAerosolveaerosolve.airbnb.techDataturedatature.ioTogether AItogether.aiOptunaoptuna.orgCatBoostcatboost.ai
Octo FAQ
What pretrained models are provided?
Octo-Base with 93M parameters and Octo-Small with 27M parameters, with reported inference rates of 13 and 17 iterations per second on one NVIDIA 4090.
How can the installation be tested?
By finetuning on the debug dataset with: python scripts/finetune.py --config.pretrained_path=hf://rail-berkeley/octo-small-1.5 --debug
What finetuning options are supported?
Three modes — head_only, head_mlp_only, and full — and three task types — image_conditioned, language_conditioned, or multimodal.
Is something wrong or out of date?
Tell us and an editor will check it.
Own this website?
Claim this listing
Claim the listing and suggest updates yourself. An editor checks every claim first.