VLAOpen2024-06-13

openvla

OpenVLA: An open-source vision-language-action model for robotic manipulation.

Best for

VLA training

Primary blocker

Verify weight files, sizes, format, and loading instructions.

Evidence rule

A model hub link is a declaration. Files, loadability, evaluation, and deployment are scored separately.

Family

VLA

Signals

5 tracked

Datasets

3 linked

Open source link Match datasets

Model decision scorecard

Use-case scores and evidence confidence are separate; Unknown is not treated as failure.

2 unresolved dimensions

Access and governance

Useful

75confidence 85

Access and license are declared by the source.

Artifact availability

Useful

55confidence 55

Source files were verified, but public weights are not available.

Training and loading reproducibility

Unknown

—confidence 15

No verified loading configuration is available.

Training data requirements

Useful

76confidence 45

Required signal categories are structured; exact tensor and action interfaces still need verification.

Evaluation evidence

Unknown

—confidence 10

No structured evaluation evidence has been verified.

Deployment readiness

Unknown

—confidence 10

Hardware, latency, dependencies, and checkpoint loading are not pipeline-tested.

Artifact facts and provenance

availabilitymetadata_verified

open

repository metadata

licensemetadata_verified

MIT

repository metadata: license

repository.filesmetadata_verified

.gitignore · .pre-commit-config.yaml · LICENSE · Makefile · README.md · experiments · experiments/robot · experiments/robot/bridge · experiments/robot/bridge/bridgev2_utils.py · experiments/robot/bridge/run_bridgev2_eval.py · experiments/robot/bridge/widowx_env.py · experiments/robot/libero · experiments/robot/libero/libero_requirements.txt · experiments/robot/libero/libero_utils.py · experiments/robot/libero/regenerate_libero_dataset.py · experiments/robot/libero/run_libero_eval.py · experiments/robot/openvla_utils.py · experiments/robot/robot_utils.py · prismatic · prismatic/__init__.py · prismatic/conf · prismatic/conf/__init__.py · prismatic/conf/datasets.py · prismatic/conf/models.py · prismatic/conf/vla.py · prismatic/extern · prismatic/extern/__init__.py · prismatic/extern/hf · prismatic/extern/hf/__init__.py · prismatic/extern/hf/configuration_prismatic.py · prismatic/extern/hf/modeling_prismatic.py · prismatic/extern/hf/processing_prismatic.py · prismatic/models · prismatic/models/__init__.py · prismatic/models/backbones · prismatic/models/backbones/__init__.py · prismatic/models/backbones/llm · prismatic/models/backbones/llm/__init__.py · prismatic/models/backbones/llm/base_llm.py · prismatic/models/backbones/llm/llama2.py · prismatic/models/backbones/llm/mistral.py · prismatic/models/backbones/llm/phi.py · prismatic/models/backbones/llm/prompting · prismatic/models/backbones/llm/prompting/__init__.py · prismatic/models/backbones/llm/prompting/base_prompter.py · prismatic/models/backbones/llm/prompting/llama2_chat_prompter.py · prismatic/models/backbones/llm/prompting/mistral_instruct_prompter.py · prismatic/models/backbones/llm/prompting/phi_prompter.py · prismatic/models/backbones/llm/prompting/vicuna_v15_prompter.py · prismatic/models/backbones/vision · prismatic/models/backbones/vision/__init__.py · prismatic/models/backbones/vision/base_vision.py · prismatic/models/backbones/vision/clip_vit.py · prismatic/models/backbones/vision/dinoclip_vit.py · prismatic/models/backbones/vision/dinosiglip_vit.py · prismatic/models/backbones/vision/dinov2_vit.py · prismatic/models/backbones/vision/in1k_vit.py · prismatic/models/backbones/vision/siglip_vit.py · prismatic/models/load.py · prismatic/models/materialize.py · prismatic/models/registry.py · prismatic/models/vlas · prismatic/models/vlas/__init__.py · prismatic/models/vlas/openvla.py · prismatic/models/vlms · prismatic/models/vlms/__init__.py · prismatic/models/vlms/base_vlm.py · prismatic/models/vlms/prismatic.py · prismatic/overwatch · prismatic/overwatch/__init__.py · prismatic/overwatch/overwatch.py · prismatic/preprocessing · prismatic/preprocessing/__init__.py · prismatic/preprocessing/datasets · prismatic/preprocessing/datasets/__init__.py · prismatic/preprocessing/datasets/datasets.py · prismatic/preprocessing/download.py · prismatic/preprocessing/materialize.py · prismatic/py.typed · prismatic/training · prismatic/training/__init__.py · prismatic/training/materialize.py · prismatic/training/metrics.py · prismatic/training/strategies · prismatic/training/strategies/__init__.py · prismatic/training/strategies/base_strategy.py · prismatic/training/strategies/ddp.py · prismatic/training/strategies/fsdp.py · prismatic/util · prismatic/util/__init__.py · prismatic/util/batching_utils.py · prismatic/util/data_utils.py · prismatic/util/nn_utils.py · prismatic/util/torch_utils.py · prismatic/vla · prismatic/vla/__init__.py · prismatic/vla/action_tokenizer.py · prismatic/vla/datasets · prismatic/vla/datasets/__init__.py · prismatic/vla/datasets/datasets.py · prismatic/vla/datasets/rlds · prismatic/vla/datasets/rlds/__init__.py · prismatic/vla/datasets/rlds/dataset.py · prismatic/vla/datasets/rlds/obs_transforms.py · prismatic/vla/datasets/rlds/oxe · prismatic/vla/datasets/rlds/oxe/__init__.py · prismatic/vla/datasets/rlds/oxe/configs.py · prismatic/vla/datasets/rlds/oxe/materialize.py · prismatic/vla/datasets/rlds/oxe/mixtures.py · prismatic/vla/datasets/rlds/oxe/transforms.py · prismatic/vla/datasets/rlds/oxe/utils · prismatic/vla/datasets/rlds/oxe/utils/droid_utils.py · prismatic/vla/datasets/rlds/traj_transforms.py · prismatic/vla/datasets/rlds/utils · prismatic/vla/datasets/rlds/utils/__init__.py · prismatic/vla/datasets/rlds/utils/data_utils.py · prismatic/vla/datasets/rlds/utils/goal_relabeling.py · prismatic/vla/datasets/rlds/utils/task_augmentation.py · prismatic/vla/materialize.py · pyproject.toml · requirements-min.txt · scripts · scripts/additional-datasets · scripts/additional-datasets/lrv_instruct.py · scripts/additional-datasets/lvis_instruct_4v.py · scripts/extern · scripts/extern/convert_prismatic_weights_to_hf.py · scripts/extern/verify_prismatic.py · scripts/generate.py · scripts/preprocess.py · scripts/pretrain.py · vla-scripts · vla-scripts/deploy.py · vla-scripts/extern · vla-scripts/extern/convert_openvla_weights_to_hf.py · vla-scripts/extern/verify_openvla.py · vla-scripts/finetune.py · vla-scripts/train.py

repository file tree

repository.codemetadata_verified

true

repository source files

dataset.signalsofficial_claim

video · actions · language · success/failure · force/tactile

README/card declaration

dataset.schemametadata_verified

true

card dataset_info or manifest

Loop signal demand

Signals this model family needs for training, evaluation, or failure mining.

6 required categories

Observation / ego video

video · video

required

Language intent / task phase

language · language · Task success across objects and scenes · Failure clusters where language grounding breaks

required

Action / robot state

actions · actions

required

Future state / dynamics

Task success across objects and scenes

required

Feedback / correction / failure

success/failure · success/failure · Task success across objects and scenes · Failure clusters where language grounding breaks

required

Sim-real / embodiment metadata

VLA

required

Evaluation focus

Task success across objects and scenes
Generalization after fine-tuning
Failure clusters where language grounding breaks

Missing critical loop signals

Core signal demands are represented. Check quality, alignment, and access constraints.

Related catalog datasets

EgoWorld

Bimanual manipulation in LeRobot format

Policy learningUseful

World modelUseful

OpenBot inference · confidence 30 · dataset readiness 79

Dataset license restricts commercial use.

EgoStation GoPro Pick-and-Place

GoPro first-person pick-and-place trajectories

Policy learningUseful

World modelUseful

OpenBot inference · confidence 45 · dataset readiness 74

Dataset license restricts commercial use.

Egocentric Adjust Bottle

Apache-2.0 LeRobot bottle-adjustment task

Policy learningUseful

World modelUseful

OpenBot inference · confidence 45 · dataset readiness 74

Exact action dimensions, control frequency, normalization, and camera mapping require interface verification.

OpenBot notes

A strong anchor for mapping dataset readiness to VLA fine-tuning needs.
Pairs naturally with Catalog filters for action/state and intent/task phase signals.

Related models