Model File Exist Error When Running on Colab

[mikenoname-github]

I have uploaded the V2-Large model to my Google Drive and am trying to call it from a Colab instance using:

!whisper "/content/drive/MyDrive/Transcription/PD3.m4a" --model large --language English --model_dir "/content/drive/MyDrive/Whisper Model/large-v2.pt" --output_format txt --condition_on_previous_text False

It results in the following output:

Traceback (most recent call last):
File "/usr/local/bin/whisper", line 8, in
sys.exit(cli())
File "/usr/local/lib/python3.10/dist-packages/whisper/transcribe.py", line 444, in cli
model = load_model(model_name, device=device, download_root=model_dir)
File "/usr/local/lib/python3.10/dist-packages/whisper/init.py", line 131, in load_model
checkpoint_file = _download(_MODELS[name], download_root, in_memory)
File "/usr/local/lib/python3.10/dist-packages/whisper/init.py", line 49, in _download
os.makedirs(root, exist_ok=True)
File "/usr/lib/python3.10/os.py", line 225, in makedirs
mkdir(name, mode)
FileExistsError: [Errno 17] File exists: '/content/drive/MyDrive/WhisperModel/large-v2.pt'

I must be doing something wrong, but can't see what it is. Can anyone help?

Thank you!

[phineas-pta]

remove --model-dir "....."

[mikenoname-github]

But then won't that defeat the purpose of my calling the model from my Drive? Or am I missing something?

[phineas-pta]

why u put the model in your drive ?

[glangford]

Agreed with @phineas-pta , it isn't really necessary to put the model on your Google Drive. You can just install it remotely each time.

That said - if for some reason you really do want to keep models on your Google Drive, there is an error in the command line. --model_dir requires a directory. So the way to do this would be

--model_dir "/content/drive/MyDrive/Whisper Model/"

(without large-v2.pt at the end)

[mikenoname-github]

Thanks Glenn, that was the piece that I missed!

And thanks Phan for your help. I always do install it remotely each time I use it, but I wanted to know if I could control the model being used without having to fetch it from a resource that I don't have control over.

[phineas-pta]

the model file is 3gb = u lost 20% of free 15gb

not sure what u want to control, the only download link for .pth file is by openai

[cod3xpert]

@mikenoname-github you have passed --model_dir "/content/drive/MyDrive/Whisper Model/large-v2.pt"
here you just need to pass model_dir which is "/content/drive/MyDrive/Whisper Model"
remove large-v2.pt it will auto pick the model file based on --model large ( model name )