# SelectCudaDevice

## Introduction

The **`SelectCudaDevice`** widget in Supervisely is a user interface element that enables users to select CUDA devices available on their agent. It automatically detects and lists the devices, displaying each with its corresponding RAM usage. Additionally, a refresh button allows users to update the list and view real-time device and memory usage. The widget also includes an event handler that activates when a user selects a device from the dropdown menu.

## Function signature

```python
SelectCudaDevice(
    get_list_on_init = True,
    sort_by_free_ram = False,
    include_cpu_option = False,
    widget_id = None,
)
```

![default](https://github.com/user-attachments/assets/4a971741-3cc8-4844-b116-6abd40c751e2)

## Parameters

| Parameters           |       Type       |                                            Description                                           |
| -------------------- | :--------------: | :----------------------------------------------------------------------------------------------: |
| `get_list_on_init`   | `Optional[bool]` | Whether to retrieve and display the list of CUDA devices upon initialization. Default is `True`. |
| `sort_by_free_ram`   | `Optional[bool]` |         Whether to sort the CUDA devices by their available free RAM. Default is `False`.        |
| `include_cpu_option` | `Optional[bool]` |            Whether to include CPU as an option in the device list. Default is `False`.           |
| `widget_id`          |       `str`      |                                  An optional ID for the widget.                                  |

### get\_list\_on\_init

Determine if you want to get the device list on widget initialization. If `True`, the widget retrieves information about device on initialization. Default is `True`.

**type:** `Optional[bool]`

**default value:** `True`

Initialize the empty widget, and then manually refresh it to find the devices:

```python
widget = SelectCudaDevice(get_list_on_init=False)

# some code here

widget.refresh()
```

### sort\_by\_free\_ram

If `True`, the device list will automatically be sorted by its free RAM value. Default is `False`.

**type:** `Optional[bool]`

**default value:** `False`

Initialize the widget sorting it by free RAM:

```python
widget = SelectCudaDevice(sort_by_free_ram=True)
```

![sort\_by\_free\_ram](https://github.com/user-attachments/assets/13f47e61-e924-4386-903a-62ae0ecb96fc)

### include\_cpu\_option

If `True`, selector will include CPU as a device option. Default is `False`.

**type:** `Optional[bool]`

**default value:** `False`

Initialize the widget with CPU as a device option:

```python
widget = SelectCudaDevice(include_cpu_option=True)
```

### widget\_id

An optional ID for the widget.

**type:** `str`

**default value:** `None`

Initialize the widget with an ID:

```python
widget = SelectCudaDevice(widget_id="select_cuda_device")
```

## Methods and attributes

|  Methods and Attributes  |                                         Description                                         |
| :----------------------: | :-----------------------------------------------------------------------------------------: |
|      `get_device()`      |                                Get currently selected device.                               |
| `set_device(value: str)` |              Set the device to be displayed in the widget (overrides current).              |
|        `refresh()`       | Get up-to-date information about available devices and set them as options in the selector. |

## Mini app example

You can find this example in our GitHub repository:

[selection/023\_select\_cuda\_device/src/main.py](https://github.com/supervisely-ecosystem/ui-widgets-demos/blob/master/selection/023_select_cuda_device/src/main.py)

### Import libraries

```python
import os

import supervisely as sly
from dotenv import load_dotenv
from supervisely.app.widgets import Card, SelectCudaDevice, Container, Text
```

### Init API client

First, we load environment variables with credentials and init API for communicating with Supervisely Instance:

```python
load_dotenv("local.env")
load_dotenv(os.path.expanduser("~/supervisely.env"))
api = sly.Api()
```

### Initialize the `SelectCudaDevice` widget

Let's initialize the `SelectCudaDevice` widget with no devices:

```python
cuda_device = SelectCudaDevice(
    get_list_on_init=False,
    include_cpu_option=True,
)
```

Then, we can manually refresh the widget, retrieving the devices information when we need to.

```python
cuda_device.refresh()
```

### Create an event handler for the `SelectCudaDevice` widget

Now, let's create a event handler for the SelectCudaDevice widget, when the value of the widget is changed.

```python
value_changed_info = Text(status="info")
value_changed_info.hide()


@cuda_device.value_changed
def new_device_selected(device: str):
    text = "Value changed: {}".format(device)
    value_changed_info.text = text
    value_changed_info.show()
```

### Create app layout

Prepare a layout for the app using `Card` widget with the `content` parameter and place widget that we've just created in the `Container` widget.

```python
card = Card(
    title="Select Cuda Device",
    content=Container(widgets=[value_changed_info, cuda_device]),
)

layout = Container(widgets=[card])
```

### Create an app using the layout

Create an app object with the layout parameter.

```python
app = sly.Application(layout=layout)
```

![mini\_app](https://github.com/user-attachments/assets/13f47e61-e924-4386-903a-62ae0ecb96fc)


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://developer.supervisely.com/app-development/widgets/selection/selectcudadevice.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.