TinyTorch v0.1.5 - Content updates and improvements

TinyTorch v0.1.5

This release includes Windows support, bug fixes, and documentation improvements.

✨ New Features

• Windows Support: Full Windows compatibility with Git Bash
  • Added PYTHONUTF8 and PYTHONIOENCODING for proper Unicode/emoji handling
  • Windows OS matrix support in CI for progressive testing

🐛 Bug Fixes

• Activations Module: Fixed Softmax forward pass implementation by @minhdang26403 in #1141
• Activations Module: Removed unnecessary Sigmoid clipping by @minhdang26403 in #1140
• Activations Module: Fixed typo and answer render error by @minhdang26403 in #1139
• Convolutions Module: Fixed computation example (Position 1,1: 8→7) - reported by @ngbolin in #1144
• Convolutions Module: Fixed pooling example element lists and averages
• Tensor Module: Fixed matrix multiplication docstring examples
• Profiling Module: Fixed convolution FLOPs calculation
• Optimizer: Fixed gradient bug and CI improvements by @profvjreddi in #1136

📝 Documentation

• Fixed broken chapter links in README by @BunningsWarehouseOfficial in #1132
• Fixed typos by @didier-durand in #1133
• Star button now links to GitHub stars explanation section

🔧 CI/Infrastructure

• Windows CI improvements (using windows-2022 for stability)
• Validate workflow now only runs on dev push, not main
• Updated workflow references to tinytorch-validate-dev

---

👥 Contributors

Thanks to all contributors who made this release possible:

• @minhdang26403
• @BunningsWarehouseOfficial
• @didier-durand
• @ngbolin
• @profvjreddi

🆕 New Contributors

• @BunningsWarehouseOfficial made their first contribution in #1132

---

Full Changelog: tinytorch-v0.1.4...tinytorch-v0.1.5

Website: https://mlsysbook.ai/tinytorch/

TinyTorch Lecture Slides v0.1.0

Initial release of TinyTorch lecture slides (PDF format) for all 18 modules.

TinyTorch v0.1.4 - Bug fixes and CI improvements

🔥 TinyTorch v0.1.4 - Bug Fixes & CI Improvements

Release Date: January 22, 2026

This release fixes test ordering issues and improves CI reliability. Note: v0.1.3 was an internal release that introduced bugs fixed here.

🐛 Bug Fixes

• Test ordering dependencies - Fixed module tests that could fail when run in different order (Fixes #1127, #1128)
• Non-interactive mode - Fixed tito dev validate --ci and tito milestone run for CI environments (Fixes #1129)
• CLI command naming - Tests now use correct command milestone (singular) instead of milestones
• Path corrections - Fixed module directory (src/ not modules/) and milestone script paths

🔧 CI Improvements

• Test summary table - CI now shows pass/fail status for each test suite in job summary
• All tests required - E2E and CLI tests must pass (removed continue-on-error)
• JUnit XML output - Test results exported for better CI integration

🧪 Test Fixes

• TransformerBlock: Use ff_dim parameter (not hidden_dim)
• LayerNorm: Use normalized_shape parameter (not embed_dim)
• Skip advanced autograd tests not supported in educational version
• Fix regression test imports to use tinytorch.core.* paths

🛠️ Other Improvements

• fix: initialize parameter gradient after creating Optimizer object by @minhdang26403 in #1114
• fix: correct RandomHorizontalFlip axis for HWC inputs by @minhdang26403 in #1113
• fix: fix module import in Transformers module test by @minhdang26403 in #1117
• fix: miscellaneous fix for Tokenizer by @minhdang26403 in #1115
• fix: fix memory calculation result by @minhdang26403 in #1118
• fix: matmul should not allow 0D tensors by @minhdang26403 in #1120
• Remove unused next_functions from Function class by @profvjreddi in #1123
• Add per-project All Contributors setup by @profvjreddi in #1125

👥 Contributors

Thanks to the contributors who made this release possible!

• @minhdang26403 - First contribution! 🎉
• @profvjreddi

📝 Notes

• No breaking API changes from v0.1.2
• Recommended update for all users, especially those running CI/CD

Learn More

• 🌐 TinyTorch Website
• 📖 Documentation
• 💬 Discussions

Full Changelog: tinytorch-v0.1.2...tinytorch-v0.1.4

TinyTorch v0.1.3 - CLI Reorganization

⚠️ TinyTorch v0.1.3 - CLI Reorganization

Release Date: January 14, 2026

This release reorganized CLI commands and added new features, but introduced test ordering bugs. Please use v0.1.4 instead.

✨ Features

• CLI reorganization: Commands restructured for better discoverability
• Welcome message: New tito bare command shows helpful welcome screen
• Feedback link: Added feedback link to announcement bar
• nbgrader docs: Marked as experimental with updated instructor guide

🐛 Bug Fixes (in this release)

• tito module complete no longer overwrites student notebooks
• Announcement bar no longer overlaps navbar

⚠️ Known Issues (Fixed in v0.1.4)

• #1127: Module 05 tests fail due to test ordering dependencies
• #1128: Module 06 autograd test ordering issue
• #1129: tito milestone run fails in non-interactive mode

📝 Upgrade Notice

Do not use this version. Please upgrade to v0.1.4 which fixes all known issues.

# Update to latest version
cd tinytorch && git pull origin main

Full Changelog: tinytorch-v0.1.2...tinytorch-v0.1.3

TinyTorch v0.1.2 - Bugfix Release

🐛 TinyTorch v0.1.2 - Bugfix Release

Stability improvements and workflow fixes.

🐛 Bug Fixes

• Workflow ordering: Integration tests now run AFTER export
• Install script: Now pulls from main branch for stability
• Announcement bar: Uses relative path for proper loading

📝 Notes

• No breaking changes from v0.1.1
• Recommended update for all users

Learn More

• 🌐 TinyTorch Website
• 💬 Discussions

TinyTorch v0.1.1 - Versioning, Audio & Polish

🚀 TinyTorch v0.1.1 - Versioning, Audio & Polish

Major update with versioning system, educational audio, and extensive polish.

✨ New Features

• Versioning system: tito --version and tito update commands
• Audio narrations: Embedded audio players in module ABOUT pages
• Educational datasets: tinydigits and tinytalks for learning
• Dismissible announcement bar: Stay informed about updates

🔧 Improvements

• Module reordering: DataLoader moved to Module 05 for better flow
• Acceleration/Memoization swap: Better pedagogical ordering (17↔18)
• Python 3.8+ support: Broader compatibility
• Progressive disclosure: Stricter import dependencies

🐛 Bug Fixes

• Fixed requires_grad issue in Linear layer Tensor calls
• Fixed navbar links from subdirectories
• SSL certificate verification on macOS
• Corrected module numbering across codebase

📚 Documentation

• Standardized ABOUT.md files for all modules
• Milestone narratives with "aha moments"
• PDF lab guide improvements

Learn More

• 🌐 TinyTorch Website
• 💬 Discussions

TinyTorch v0.1.0 - Initial Public Release

🎉 TinyTorch v0.1.0 - Initial Public Release

First public release of TinyTorch educational ML framework.

Features

• 📚 20 progressive modules: Build from Tensor to Transformer
• 🛠️ tito CLI: Guided learning experience
• 🏆 6 milestone projects: Real-world applications
• ✅ Comprehensive test suite: Validate your implementations
• 📖 Jupyter Book documentation: Interactive learning

Getting Started

curl -fsSL https://mlsysbook.ai/tinytorch/install.sh | bash

Learn More

• 🌐 TinyTorch Website
• 📖 Getting Started Guide
• 💬 Discussions

book-v0.5.1: Content updates and improvements

Release v0.5.1: Illuminating Machine Learning Systems Through Clarity and Depth

This patch release of MLSysBook focuses on refining the learning experience by enhancing clarity, depth, and accessibility for all users – students, researchers, practitioners, educators, and contributors alike. We've meticulously reviewed content, visualizations, and technical aspects to deliver a more engaging and impactful exploration of machine learning systems.

✨ Major Features

📖 Content Improvements

• Enhanced Visualizations: Key concepts are now illustrated with updated visuals that offer improved clarity and understanding. Complex algorithms and system architectures are rendered in a more intuitive manner, aiding comprehension and retention.
• Streamlined Mathematical Notation: We've carefully reviewed and refined mathematical expressions throughout the textbook to ensure they are concise, accessible, and aligned with modern best practices. This fosters a smoother learning experience for readers with diverse mathematical backgrounds.
• Refined Explanations: Key concepts have been re-articulated with greater emphasis on pedagogical clarity, ensuring a deeper understanding of fundamental principles and advanced topics.

🛠️ Technical Excellence

• Improved Accessibility Features: We've implemented enhancements to ensure the textbook is more inclusive for all learners. This includes updated features for screen readers, improved color contrast, and alternative text descriptions for images, promoting accessibility and engagement for diverse users.
• Faster Build Process: Behind-the-scenes optimizations have resulted in a faster build process, enabling quicker access to the latest content updates and contributing to a smoother user experience.

🎓 Educational Innovation

• Interactive Learning Prompts: New interactive learning prompts have been integrated throughout the textbook, encouraging active engagement with the material and promoting deeper comprehension. These prompts facilitate critical thinking and self-assessment, enhancing the overall learning journey.
• Real-World Application Examples: Practical applications of machine learning systems are now further emphasized through the inclusion of real-world case studies and examples. This connection to practical use cases strengthens the relevance and applicability of the material for students and practitioners.

🌟 Key Achievements

For Students: Enhanced visualizations, streamlined mathematical notation, and interactive learning prompts foster a deeper understanding of complex concepts.
For Educators: New teaching resources, accessible features, and real-world application examples enrich classroom instruction and engagement.
For Contributors: A streamlined development workflow and improved documentation facilitate active participation in the open-source project.

🔬 Educational Impact

These changes cultivate a more engaging, accessible, and impactful learning experience for all users of MLSysBook. The enhanced clarity, depth, and interactivity promote a deeper understanding of machine learning systems, empowering students, researchers, practitioners, and educators to confidently navigate this rapidly evolving field.

🌐 Access Your Enhanced Textbook

• 📖 Online Version: mlsysbook.ai
• 📄 PDF Download: Available from release assets
• 📚 EPUB Version: Available from release assets
• 🧪 Labs & Exercises: Hands-on learning materials

📞 Community & Contributions

We extend our sincere gratitude to the educators, students, and practitioners who have contributed their valuable feedback. Your insights have been instrumental in shaping this release. We encourage continued engagement through our GitHub repository.

---

Development Period: [Timeframe based on release type]
Repository: harvard-edge/cs249r_book
Focus: [Main theme of this release]

---

Full Change Log

Release Notes for v0.5.1

Changes since v0.5.0

Recent Commits:

b8510b9 Merge branch 'dev'

TinyTorch Audio v0.1.1

TinyTorch Module Audio Introductions

Audio introductions for all 20 TinyTorch modules, generated using NotebookLM.

Each ~1:30 minute audio provides:

• Module concept introduction
• Why building from scratch matters
• Systems insights preview

Files

01_tensor.mp3 through 20_capstone.mp3

Usage

Embedded on TinyTorch module landing pages at mlsysbook.ai/tinytorch.

book-v0.5.0: The TinyTorch Release 🔥

Release Date: December 14, 2025

The highlight of this release is the public preview of Tiny🔥Torch, the companion hands-on learning platform integrated into the MLSysBook repository as a monorepo. This release represents a major milestone in providing students with a complete "build your own PyTorch from scratch" experience.

---

🔥 TinyTorch Platform

• Public Preview Release: TinyTorch officially launches as the hands-on companion to the MLSysBook, allowing students to build their own deep learning framework from the ground up. Includes 20 progressive modules from tensors to transformers.
• Module Documentation: Comprehensive ABOUT.md files generated for all 19 modules (02-20), providing standardized pedagogical documentation including learning objectives, module connections, and "aha moment" descriptions.
• Milestone System: Major overhaul including the new consolidated transformer attention milestone, rewritten XOR crisis milestone for clarity, and CIFAR-10 Part 2 with DataLoader showcase.
• PDF Lab Guide: Complete PDF documentation with improved admonition colors, mermaid diagram sizing, TOC depth settings, and TeX Gyre fonts.
• Tito CLI: Enhanced CLI with milestone name aliases, module view command, centralized color theme system, and improved health/status commands.
• Branding & Styling: Standardized "Tiny🔥Torch" branding across the codebase with consistent emoji placement and fire theme.

📖 Book Updates

• Repository Restructure: Book content reorganized under book/ directory to accommodate the monorepo structure with TinyTorch.
• Improved Figures: Enhanced figure captions with bold titles and descriptions.
• Website: Updated announcement banner celebrating 10K GitHub stars, subscribe modal improvements.

🔧 Infrastructure

• Monorepo Integration: TinyTorch integrated into the MLSysBook repository with dedicated CI/CD workflows and the Tito CLI tool.
• CI/CD Improvements: Added preflight checks before dev and live deploys, synced PDF build between Makefile and CI workflow.
• Code Quality: Pre-commit hooks, codespell configuration updates, and consistent formatting.

---

🌐 Access

• 📖 Online Version: mlsysbook.ai
• 🔥 TinyTorch: mlsysbook.ai/tinytorch
• 📄 PDF Download: Available from release assets
• 📚 EPUB Version: Available from release assets

📋 Release Information

• Previous Version: v0.4.2
• Type: Minor Release (Major Feature Addition)
• Classroom Ready: Summer/Fall 2026

---

This release marks a significant expansion of the MLSysBook ecosystem, providing students with both the theoretical foundation (textbook) and hands-on implementation experience (TinyTorch) for understanding machine learning systems.