> ## Documentation Index
> Fetch the complete documentation index at: https://docs.orchestrallabs.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# InternW0-Delta

> Inspect and deploy the public InternW0-Delta LIBERO policy.

## Checkpoint

Inspect the checkpoint to see its camera and action contract:

```bash title="Your computer" sample=verified theme={null}
orchestral models inspect internw0_delta hf://InternRobotics/InternW0-Delta-Libero@00524e29e160b2f092bf6c5a47e325301a31fb90
```

```text title="Sample output" theme={null}
Inputs:
  observation.images.agentview_image: RGB uint8 (256, 256, 3)
  observation.images.robot0_eye_in_hand_image: RGB uint8 (256, 256, 3)
  state: float32 (16,)

Outputs:
  actions: float32 (32,7)
```

## Inputs and output

Provide both RGB camera frames and the 16-value LIBERO state vector. The policy returns a chunk
of 32 seven-value actions.

| Direction | Name | Shape | Type |
| - | - | - | - |
| Input | `observation.images.agentview_image` | `(256, 256, 3)` | RGB `uint8` |
| Input | `observation.images.robot0_eye_in_hand_image` | `(256, 256, 3)` | RGB `uint8` |
| Input | `state` | `(16,)` | `float32` |
| Output | `actions` | `(32, 7)` | `float32` |

## Deploy and predict

Create an endpoint from the checkpoint URI. The default deployment uses 10 inference steps and
the request-graph execution profile.

```bash your computer theme={null}
orchestral deployments create hf://InternRobotics/InternW0-Delta-Libero@00524e29e160b2f092bf6c5a47e325301a31fb90 --robot robot-1 --wait
```

```text title="Sample output" theme={null}
calm-violet-otter-j9q2  ACTIVE · ready
InternW0-Delta · robot-1
Timeline: requested 12s ago → starting worker 8s ago → ready now
```

```python robot computer theme={null}
import numpy as np
import orchestral as ol

client = ol.Orchestral()
deployment_id = input("Deployment ID from the CLI output: ").strip()
deployment = client.deployments.get(deployment_id)

inputs = {
    "observation.images.agentview_image": np.zeros((256, 256, 3), dtype=np.uint8),
    "observation.images.robot0_eye_in_hand_image": np.zeros((256, 256, 3), dtype=np.uint8),
    "state": np.zeros((16,), dtype=np.float32),
}

with deployment.session() as session:
    result = session.predict(inputs=inputs, instruction="pick up the object")

print(result.actions.shape)  # (32, 7)
```

## Checkpoint identification

When importing a renamed copy of the checkpoint, Orchestral identifies the family from the
serialized model metadata in the weight archive. Detection does not depend on the Hugging Face
owner or repository name. The adapter validates the inputs and outputs before loading the policy.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.