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

# cassian up

> Start a GPU instance.

## Usage

```bash theme={null}
cassian up [options]
```

Reads `cassian.yaml`, picks the best available node, starts your container, and restores your workspace.

If an instance with the same name is already running, it stops it first.

## Options

| Flag           | Description                                                 |
| -------------- | ----------------------------------------------------------- |
| `--gpu <type>` | Override GPU type from cassian.yaml (e.g. `l4`, `a10g`)     |
| `--no-sync`    | Skip pushing local files after the instance starts          |
| `--no-setup`   | Skip `workspace.setup` and auto-detected dependency install |

## What happens

1. Validates your `cassian.yaml` (including GPU type check)
2. Connects to Cassian and provisions a container
3. Restores your workspace files from the last session
4. Prints runtime info (Python version, Torch version, GPU model)
5. Installs dependencies (auto-detected or from `workspace.setup`)
6. Shows connection instructions

## Startup time

First `cassian up` for a new volume takes 30 to 60 seconds depending on how much data is in your workspace.

Repeat `cassian up` (after `cassian down`) is faster because your volume is cached on the node. Typically under 20 seconds.

## Switching GPUs without editing cassian.yaml

```bash theme={null}
cassian up --gpu l4
cassian up --gpu a10g
```

Useful when your preferred type is unavailable. The flag overrides `gpu.type` for this session only — your cassian.yaml is unchanged.

## Fast restart

```bash theme={null}
cassian up --no-sync --no-setup
```

Skips both the file push and the setup command. Use when your environment is already installed and your files haven't changed — shaves 30+ seconds off restarts.

## Auto dependency install

If `workspace.setup` is not set in `cassian.yaml`, Cassian detects and installs automatically:

* **`requirements.txt`** in project root runs `pip install -r requirements.txt`
* **`pyproject.toml`** in project root runs `pip install -e .`

To disable this: `cassian up --no-setup`

To set an explicit command:

```yaml theme={null}
workspace:
  setup: "pip install -r requirements.txt && python setup_data.py"
```

## GPU type validation

```
x Unknown GPU type 'rtx-3090'.
  Did you mean: rtx3090?
```

## Runtime banner

```
  Python      3.11.9
  Torch       2.6.0+cu124
  GPU         NVIDIA GeForce RTX 3090
  Image       default
```