-
Notifications
You must be signed in to change notification settings - Fork 72
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
the method get_onnx_model() should not need the path to original model #38
Comments
yes, it is bit confusing, it should be the purpose of the
can you elaborate more on this. |
I agree. This is exactly my point: the name of the model is important to get back our ONNX model with My second point is that
I think we could save the |
Hi,
After having obtained our ONNX model thanks to the
export_and_get_onnx_model()
method, we want to load it in order to use it in production.There is a method called
get_onnx_model()
which requires the following arguments:model_name_or_path
,onnx_models_path=saved_models_path
andquantized=True
This is conceptually strange: why do I need to access the original model (via
model_name_or_path
) when I want to use the ONNX one?I explored the code and figured out that
model_name_or_path
is needed for 2 things:In order to avoid passing the
model_name_or_path
argument inget_onnx_model()
, theexport_and_get_onnx_model()
method could save the model name configuration in the ONNX folder and could arrange a standard model name for ONNX files names as we can now (in the latest version of fastt5) customize ONNX folder path.What do you think?
The text was updated successfully, but these errors were encountered: