[BUG]: Audio ingestion does not automatically use a GPU resource

### Version

26.03

### Which installation method(s) does this occur on?

_No response_

### Describe the bug.

I should be able to use the snippet for audio ingestion in the quickstart using the Python API only, and it should automatically use a GPU for the parakeet model.  

### Minimum reproducible example

```shell

```

### Relevant log output

```shell

```

### Other/Misc.

_No response_