Deep Learning: From Theory to Practice

Practical assignments from a Deep Learning course (MAP583, École Polytechnique — based on the dataflowr material), implemented in PyTorch.

Contents

Notebook	Topic
TP00	Python warm-up
TP01	MLP on a toy dataset
TP02	Writing a full training loop
TP03	Optimization algorithms (SGD, momentum, Adam…)
TP04	Convolutional networks
TP05	Sequence prediction with RNNs
TP06	Stability and robustness of training
TP07	Language modeling

The Lesson/ folder contains complementary notebooks on RNNs and packed sequences.