Vijay Janapa Reddi
|
3fc83f95d6
|
Fix tutorial master plan: Logical module sequence for Phase 2
- Phase 2 now: Autograd → Spatial → Optimizers → DataLoader → Training
- Move Spatial (CNNs) from Phase 3 to Phase 2 Module 7
- Integrate BatchNorm into Spatial module (mirrors PyTorch patterns)
- Fix milestone: CNN training achievable at end of Phase 2 (Module 10)
- Phase 3 focuses on language: Tokenization → Embeddings → Attention → Transformers
- Logical dependency flow: understand conv operations before optimizing them
|
2025-09-23 18:28:44 -04:00 |
|