Log if handler service is using default or custom functions implementation

[aduverger]

Issue #, if available:

• HF_TASK Enviournment Variable error #110
• Sagemaker endpoint inferencing error with HF model loading from s3bucket with new transformer update #108
• https://discuss.huggingface.co/t/inference-failed-for-flan-ul2-20b-on-sagemaker/35359
• https://discuss.huggingface.co/t/hf-task-enviournment-variable-error/68927

Description of changes:
Currently, when there is no model_fn function implemented by the user (or when the handler service fails to load such implementation), the default load function is being used to load the model.

Because it's not clear from the logs when that default function is being used, this makes things difficult for users to debug when they were expecting their own implementations to be used.
E.g.: let's say I write a inference.py script with a model_fn. For some reason, this function is not being picked up by the handler service. This reason could be:

• The name of the method is not exactly model_fn
• An issue with the archive file: e.g. the archive is named model_artifacts_v2.tar.gz, but the root folder in this archive is named model_artifacts and not model_artifacts_v2
• ...

Because the default load function is being used in that case, the logs would display an error such as:

 "You need to define one of the following [...] as env 'HF_TASK'.",

Which make things confusing as the user thinks they need to set a env variable to fix the issue, whereas the problem is related to the custom implementation not being used.

This PR adds some logs to tell the user which implementation are being used for each function.

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[aduverger]

Hey @philschmid, any chance you could take a look at this?
Thanks!

[aduverger]

Anyone around 😅?

[philschmid]

Sorry, took a while. Addded it. Its never bad to add some extra logs.