I need to make an AI model in Python from scratch, untrained but with a view to being fine-tuned with my dataset later. I have some experience with machine learning frameworks like TensorFlow and PyTorch but don’t know how to get started with a blank model that might be fine-tuned.

Specifically, I need guidance on the following.

How to create a very basic, non-trainable model in Python.
How to structure the model to support fine-tuning.
Best practices for loading and pre-processing my dataset for training. Steps for training the model on my dataset:.

Here is an outline of what I have tried so far:

Installed TensorFlow and PyTorch. Further, simple neural network models were created, but they were pre-trained.

I would appreciate any examples, code snippets, or references to relevant documentation that could help me achieve this. Thank you!