> ## 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.

# ABC-VLA

> Inspect and deploy ABC-VLA for three-camera bimanual observations and action chunks.

## Checkpoint

This guide uses the pinned ABC-VLA Jetson Thor checkpoint. Inspect it before deployment to
confirm the camera and state contract:

```bash title="Your computer" sample=verified theme={null}
orchestral models inspect abc_vla hf://agents2agents/ABC-VLA-Jetson-Thor@c2f154f31115925e61d626da6d900d9f2089590e
```

```text title="Sample output" theme={null}
Inputs:
  top: RGB uint8 (224, 224, 3)
  left: RGB uint8 (224, 224, 3)
  right: RGB uint8 (224, 224, 3)
  state: float32 (14,)

Outputs:
  actions: float32 (30,14)
```

## Inputs and output

Provide all three camera views and the 14-value state vector. Missing or extra input keys raise
an error.

| Direction | Name | Shape | Type |
| - | - | - | - |
| Input | `top` | `(224, 224, 3)` | RGB `uint8` |
| Input | `left` | `(224, 224, 3)` | RGB `uint8` |
| Input | `right` | `(224, 224, 3)` | RGB `uint8` |
| Input | `state` | `(14,)` | `float32` |
| Output | `actions` | `(30,14)` | `float32` |

Each prediction returns a 30-step action chunk. Use the checkpoint's expected robot state and
action convention when connecting the output to a robot controller.

## Deploy and predict

Deploy the pinned checkpoint directly. Orchestral resolves and prepares the model before the
serving deployment becomes ready.

```bash your computer theme={null}
orchestral deployments create hf://agents2agents/ABC-VLA-Jetson-Thor@c2f154f31115925e61d626da6d900d9f2089590e --robot robot-1 --wait
```

```text title="Sample output" theme={null}
jolly-tangerine-snail-rdik  ACTIVE · ready
ABC-VLA · 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 = {
    "top": np.zeros((224, 224, 3), dtype=np.uint8),
    "left": np.zeros((224, 224, 3), dtype=np.uint8),
    "right": np.zeros((224, 224, 3), dtype=np.uint8),
    "state": np.zeros((14,), dtype=np.float32),
}

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

print(result.actions.shape)  # (30, 14)
```


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