Transformer/Pipelines Tutorial - where is it running?

Beginners
1

Hey HuggingFace Experts - thank you for all the work you do supporting the community. I’m a noob running the Transformers tutorial, got to the pipeline tutorial here Pipelines for inference.

Through my terminal, I did the pip install within a virtual environment (.env), then pip install transformers, and then torch.

I then moved on to the MLK example, successfully running it with wav2vec2 and then whisper-large - awesome!

I also successfully used it to transcribe some local mp3 files - awesome! awesome!

When following the instructions to set up whisper, it seemed to auto download a 6.7GB file from HF to my machine.

Q’s

  • What exactly is this file i downloaded? the model itself, the weights?
  • Is this file saved on my machine persistently, does it get trashed when i shut down the terminal and virtual environment (.env)?
  • if so, any help finding the file?
  • am I running this model locally on my CPU/GPU or on HuggingFace CPU/GPU?

Thank you for the assitance.

2

Hi there!

  • You’re downloading the files here required to run the model - weights, configurations, etc.
  • The files are saved to your cache ( ~/.cache/huggingface/hub), you can read more about it here if you want to save your files somewhere else.
  • The model is running locally on your hardware :slight_smile:
3

Thank you very much for taking the time to answer! :hugs:

4

could you explain a bit for how to assess the local hardware spec for running the model before download such size model file?
I reviewed the model card tab but see nothing directly about it. or just simply caculate based on the Model size of params?
As a individual developer, the RTX gaming GPU is the popular choice I believe.