When deploying a model, and selecting the dataconnection, some labs state to only enter the path without the actual model's filename (eg. chapter11s04 models) and soms labs state to enter the path including the model's filename (eg chapter11s02 models/rf_iris.onnx).
Both seem to work fine.
So it seems the runtime tries to find the correct onnx file automatically if you only enter the foldername, but isn't that risky?
@IvanLabrovic this is related with how the models are packaged. Some models are not packaged in a single file and provide the model file, the weights, and the configuration in different files in a directory. Others, such as ONNX or GUFF, are selfcontained and everything is inside one file.
@IvanLabrovic this is related with how the models are packaged. Some models are not packaged in a single file and provide the model file, the weights, and the configuration in different files in a directory. Others, such as ONNX or GUFF, are selfcontained and everything is inside one file.
Red Hat
Learning Community
A collaborative learning environment, enabling open source skill development.