cs249r_book/tinytorch/CHANGELOG.md

Changelog

All notable changes to TinyTorch will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

Unreleased

[0.1.5] - 2026-01-27

Added

• Windows/Git Bash support for installer script (thanks @rnjema, @joeswagson, @Kobra299!)
• Platform detection (get_platform()) for OS-specific guidance
• Cross-platform line endings via .gitattributes

Changed

• Installer uses $PYTHON_CMD -m pip for more reliable pip invocation
• Dynamic tito --version command showing current TinyTorch version

Fixed

• Virtual environment activation now works correctly on Windows Git Bash

0.1.1 - 2025-01-13

Fixed

• Module 03 (Layers): Removed premature requires_grad from Linear layer initialization
  • Aligns with progressive disclosure model where requires_grad is introduced in Module 06
  • Fixes issue where students running modules in sequence encountered undefined parameters

Added

• tinydigits dataset: 8x8 handwritten digits for educational CNN training
• tinytalks dataset: Q&A pairs for transformer training examples

0.1.0 - 2024-12-12

Added

• Initial public release of TinyTorch
• 20 progressive modules covering ML fundamentals to advanced topics
• tito CLI for guided learning experience
• Milestone projects demonstrating historical ML breakthroughs
• Comprehensive test suite
• Jupyter Book documentation site

Modules

• 01: Tensor (NumPy wrapper with ML semantics)
• 02: Activations (Sigmoid, ReLU, Tanh, GELU, Softmax)
• 03: Layers (Linear, Dropout)
• 04: Losses (MSE, CrossEntropy)
• 05: DataLoader (batching, shuffling)
• 06: Autograd (automatic differentiation)
• 07: Optimizers (SGD, Adam)
• 08: Training (training loops)
• 09: Convolutions (Conv2D, pooling)
• 10: Tokenization (BPE, character level)
• 11: Embeddings (word embeddings)
• 12: Attention (self attention, multi head)
• 13: Transformers (encoder, decoder)
• 14: Profiling (timing, memory)
• 15: Quantization (INT8, dynamic)
• 16: Compression (pruning, distillation)
• 17: Acceleration (SIMD, parallelism)
• 18: Memoization (caching, checkpointing)
• 19: Benchmarking (MLPerf style)
• 20: Capstone (integration project)