the method get_onnx_model() should not need the path to original model about fastt5

[Ki6an]

yes, it is bit confusing, it should be model_name, not model_name_or_path, I'll make this change in the next update.

the purpose of the model_name (the current model_name_or_path) is to select a particular model from a custom folder if you've stored more than one type of model in that folder.

the export_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.

can you elaborate more on this.

[piegu]

it should be model_name, not model_name_or_path

I agree. This is exactly my point: the name of the model is important to get back our ONNX model with get_onnx_model(), not the path to the model.

My second point is that get_onnx_model() calls the class OnnxT5 in step 4, and this class is initialized with the following code:

config = AutoConfig.from_pretrained(
            model_or_model_path, use_auth_token=get_auth_token()
        )

I think we could save the config file through export_and_get_onnx_model() . If not, we need model_or_model_path as argument of get_onnx_model() but we do not want that.