How to enable GPU during transcription in Python rather than in shell?

[FYNIXqwq]

I'm planning to write a program in Python letting me drag the file and starts the transcription immediately by GPU. This is convenient in shell by --device cuda. However, I could not find a detailed guide of utilization in Python.

[FYNIXqwq]

I'm trying

model = whisper.load_model("medium")
result = model.transcribe("100.mp4", fp16 = False)
print(result["text"])

and the program starts to run on CPU, which is slow.

[phineas-pta]

whisper.load_model(..., device="cuda")