Learning AI/ML

• AI for Everyone course
• Practical Deep Learning for Coders and the Jupyter notebook based book
• Huggingface
• FastAPI Docker example
• https://karpathy.ai/zero-to-hero.html

Notes

Model is a neural network function.

Models alone aren't good enough, they need to be trained and importantly you need the "weights" in order to have a ready to use model like the resnet18 model that "comes with" FastAI.

Fine tuning is taking a trained model and training it further on your data.

Learner is model + training data

Weights are parameters (hmm followup on this)

• It used to be Input > Program > Output
• Now its Input + Weights > Model > Output
• With a training loop is Input + Weights + "Loss" from previous training > Model > Output + Loss (now do it all again N times to solve any computable function!)