When I save fine-tuned weight, the weight size is doubled why?

[kouohhashi]

I fine-tuned a model and saved it like below.

# load model 
model = whisper.load_model("base")

# do some fine-tuning...

# save weights
torch.save({
    "dims": dims,
    "model_state_dict": model.state_dict(),
}, "test.pt")

The original file size was 139M.
But the file size of saved weights was 278M.
I'm confused.

How can I fix this?
I want the fine-tuned model to be 139M too.

Thanks in advance.

[kouohhashi]

I got it.
I think what I should do is model = model.half() before saving weights.