Embodied · vla · vla.policy_rollout · robot_policy · cross_embodiment_policy

Being-H0.5

BeingBeyond

WorldFoundry keeps the inference-only Being-H0.5 architecture and policy runtime directly under worldfoundry/synthesis/action_generation/being_h05.

Runner verifiedDedicated environmentbeing-h05

Build your run

Choose a recorded variant, then copy the exact setup or inspection command.

worldfoundry.pipeline

Variant

Taskvla.policy_rollout

Environmentworldfoundry-being-h05

DeviceCu118 Or Newer

worldfoundry-eval evaluate \
  --mode model \
  --model-id being-h05-2b \
  --model-runner worldfoundry:pipeline \
  --model-manifest-dir worldfoundry/data/models/catalog \
  --requests-path tmp/requests.jsonl \
  --output-dir tmp/model_eval/being-h05-2b \
  --metric artifact_count \
  --json

Compatibility & versions

Manifest-backed recipe

Integration: integrated
Runner evidence: verified
Environment: worldfoundry-being-h05
Python: 3.10
CUDA: Cu118 Or Newer
PyTorch: torch
Source revision: —
Checkpoint revision: —

Runtime profile: being-h05
Pipeline binding: being-h05
Runner: worldfoundry.pipeline
Pipeline target: worldfoundry.pipelines.component_pipelines:BeingH05Pipeline
Backend stage: in_tree_runtime
Runtime status: in_tree_being_h05_libero_get_action_gpu_ready_efficient_sdpa
Driver status: checkpoint_runtime_ready_memory_depends_on_variant
Environment kind: Dedicated environment

Install environment

The environment resolver reads the recorded profile and chooses the unified or dedicated environment shown here.

bash scripts/setup/model_env_install.sh --model being-h05-2b

Package constraints 11

numpy
Pillow
torch
torchvision
timm
numpydantic
pipablepytorch3d
safetensors
transformers
einops
pydantic>=2

Conda packages 2

python
pip

Checkpoints & assets

Run the local check before allocating compute. Gated, private, and license fields below come directly from the checkpoint manifest.

worldfoundry-eval zoo model-download --model-id being-h05 --check-local --json

BeingBeyond/Being-H05-2B

Revision: —
License: —
Gated: —
Private: —

BeingBeyond/Being-H05-2B_libero

Revision: —
License: —
Gated: —
Private: —

BeingBeyond/Being-H05-2B_robocasa

Revision: —
License: —
Gated: —
Private: —

BeingBeyond/Being-H05-2B_libero_robocasa

Revision: —
License: —
Gated: —
Private: —

Launch & outputs

The generated command uses the shared evaluation boundary and writes durable result manifests and artifacts.

worldfoundry-eval evaluate \
  --mode model \
  --model-id being-h05-2b \
  --model-runner worldfoundry:pipeline \
  --model-manifest-dir worldfoundry/data/models/catalog \
  --requests-path tmp/requests.jsonl \
  --output-dir tmp/model_eval/being-h05-2b \
  --metric artifact_count \
  --json

Input contract

Field	Recorded contract
`actions`	robot_action, action_unified, proprio, action_space
`data_config_name`	libero_nonorm
`dataset_name`	libero_posttrain
`embodiment_tag`	libero
`image`	Required
`instruction`	Required
`prompt`	Required
`state`	Required
`video`	Optional

Artifact contract

Artifact kind	Filename / path
`action_trace`	`—`
`action_trace`	`being_h05_action_trace.json`

Evidence & sources

Catalog integration, native-demo parity, and runner parity are independent records.

Integration: integrated
Runner evidence: verified
Native demo evidence: Not recorded
Validation imports: numpy, PIL, torch, torchvision, transformers, safetensors, pytorch3d.transforms

Recipe provenance

worldfoundry/data/models/catalog/vla_va_wam/being-h05.yamlGitHubhttps://github.com/BeingBeyond/Being-HHugging Facehttps://huggingface.co/BeingBeyond/Being-H05-2BHugging Facehttps://huggingface.co/BeingBeyond/Being-H05-2B_liberoHugging Facehttps://huggingface.co/BeingBeyond/Being-H05-2B_robocasaHugging Facehttps://huggingface.co/BeingBeyond/Being-H05-2B_libero_robocasa

Build your run

Input contract

Artifact contract

Related recipes