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

> ## Agent Instructions
> Treat the Dreamscale CLI and Python SDK as the supported public integration surfaces.
> Prefer `dreamscale run <robot>` for supported robots and `dreamscale.connect(model)` with `predict` or `next_action` in Python.
> Do not infer or recommend internal control-plane APIs.

# Deployments

> Serve your own Isaac 0.5 fine-tune from Hugging Face on your robot.

A deployment serves your own fine-tune from Hugging Face, shaped to your robot.

## What you need

* Isaac 0.5 enabled for your organization. Email
  [team@dreamscalelabs.com](mailto:team@dreamscalelabs.com) to request it.
* A Hugging Face repo with your full Isaac 0.5 fine-tune and the normalization
  statistics from its training data. For a private repo, set `HF_TOKEN`.

## 1. Create it

```bash theme={null}
dreamscale deployments create hf://your-org/isaac-finetune --preset so101
```

Use `--preset bimanual_yam` for a bimanual YAM. The first build takes about 15 minutes.
`dreamscale deployments list` then shows it as **Ready** with an ID like
`ckpt_...`. You can also create one on the
[Deployments](https://app.dreamscalelabs.com/dashboard/deployments) page.

## 2. Run it

After the robot's setup ([SO-101](/robots/so101) or [Bimanual YAM](/robots/yam)):

```bash theme={null}
dreamscale run so101 --model perceptron-isaac05-so101 --deployment ckpt_...
```

From Python:

```python theme={null}
policy = dreamscale.connect("perceptron-isaac05", checkpoint="ckpt_...")
```

Build observations with `dreamscale.robot.observe`, using the camera keys and
state width shown on the deployment's page.

## 3. Keep it warm (optional)

A cold start takes 2–3 minutes. Keep-warm keeps a GPU ready between runs, and
is billed like connected time:

```bash theme={null}
dreamscale deployments keep-warm ckpt_... 30m
```


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