PyTorch

What Are We Actually Doing?
The Tensor: What Is Actually Flowing Through the Loop?
Autograd: What Did PyTorch Record, and Where Does the Gradient Stop?
The Network: What Is It Without nn.Module?
nn.Module: What Does PyTorch Think Belongs to Your Model?
DataLoader: Where Is the Training Loop Actually Waiting?
Transforms: What Does the Model Actually See?
CNN Geometry: What Shape Reaches the Next Layer?
Feature Space: What Does a Linear Model Actually See?
Attention: Which Position Is Comparing With Which?
Training: Which Link in the Learning Chain Is Broken?
Performance: What Is the Machine Waiting For?
Compilation: Which Assumption Stopped Holding?
Regressions: Did the Model Change, or the Measurement?
Assembly: A Language Model You Can Interrogate
Appendix A: PyTorch Diagnostic Field Guide
Make the Automaton Differentiable
Learn the Local Update Rule
Hidden Cell Channels and Local Memory
Grow a Target From One Seed
Randomize the Update Schedule
Train for Persistence
Regenerate After Damage
Test Generalization Beyond Training
Neural Cellular Automata for Pathfinding
Learn to Solve Mazes
Generalize to Harder and Larger Mazes
Inspect Hidden-State Propagation
What Did the Neural CA Actually Learn?
Run Cellular Automata on the GPU
Preference Rankers — Learning Which Answer Is Better
Which Model Should You Use? MR.Q, EBT, SICQL, HRM, Tiny and PACS Compared
PACS — Building an Optimizer From Gradient Statistics
Inside Tiny — Residual Blocks, Attention and Sparse Autoencoders
Tiny — Recursive Reasoning With a Small Neural Network
HRM — Hierarchical Reasoning With Fast and Slow Recurrent State
SICQL — Building a Model From Q, V and Policy Networks
EBT — From One Score to Q, V, Policy and Advantage
MR.Q — Building a Neural Quality Model From Two Embeddings
The Model Inside the Model
MR.Q: Model-Based Representations for Model-Free Trading
Writing Neural Networks with PyTorch