Running from Jupyternote

[him1532]

I am trying to run your example in jupyternotebook.
I've installed all the relevant library/module

My first question is when I try to import the WhipserModel
I get
ModuleNotFoundError: No module named 'faster_whisper'

When I run this on a terminal this step is fine. Is there a way to solve this ?

My second question is regarding the compute_type parameter.
When I try the float16 I get the following error
ValueError: Requested float16 compute type, but the target device or backend do not support efficient float16computation.

I get the same error for int8_float16. So how do you know which is proper parameter before you run into error?
Does this mean I can't use cuda? I have the cuda library installed.

Thank you.

[guillaumekln]

I'm not familiar with Jupyter Notebook so I can't really help for the installation issue in this environment.

Regarding the error with compute_type, it means your GPU does not support float16 execution. You can either remove the argument entirely or set other compute types such as compute_type="float32" or compute_type="int8".

[otakutyrant]

OMG! My computer can run whisper with small.en well but faster-whisper cannot work with the same model for me until I change the compute_type from float16 to float32.