Repository navigation
Releases
v1.1.0 - Improved Broadcasting & Slicing
✨ Feature Release
This release addresses two major limitations from v1.0.0 and enhances core tensor operations.
🔧 Fixed
Broadcasting Gradient Reduction - Fixed incorrect gradient computation when broadcasting occurs in forward pass:
add_backward(): Now correctly reduces gradients when shapes differ (e.g.,[3,1] + [3,4])mul_backward(): Now correctly reduces gradients when shapes differ (e.g.,[3,1] * [3,4])matmul_backward(): Fixed two critical bugs:- Transpose now only swaps matrix dimensions (last 2), not batch dimensions
- Added gradient reduction for batch broadcasting (e.g.,
[1,3,4] @ [2,4,5])
- Added
reduce_grad_for_broadcast()helper function
⭐ Added
Enhanced Slice Functionality - Full NumPy-compatible slicing with reverse support:
tofu_tensor_arange(): Now supports negative steps for reverse slicingtofu_tensor_rearange(): Now supports negative steps for reverse slicing- Examples:
arange(10, 0, -1)→[10, 9, 8, ..., 1]✅arange(5, 5, 1)→ NULL (empty range) ✅arange(0, 10, -1)→ NULL (incompatible step) ✅
Comprehensive Test Suites:
test/standalone/test_broadcast_gradient.c: 5 broadcast gradient teststest/standalone/test_arange_enhanced.c: 6 test categories with 18+ tests
📊 Testing
- ✅ All 66 unit tests pass
- ✅ Phase 1 validation (gradient correctness)
- ✅ Phase 2 validation (architecture)
- ✅ Phase 3 validation (classification & regression)
- ✅ 23+ new tests for enhanced functionality
🚀 Performance
- No performance regressions
- Minimal overhead for gradient reduction (only when broadcasting occurs)
📚 Documentation
- Updated README with v1.1.0 announcement
- Enhanced mdBook user guide with reverse slicing examples
- Updated Doxygen API comments for arange/rearange functions
- Comprehensive CHANGELOG entry
🔗 Links
Upgrade Note: This release is fully backward compatible with v1.0.0. No breaking changes.
v1.0.0 - First Stable Release 🎉
This is the first stable release of Tofu with a frozen public API.
🔒 API Stability Guarantee
All public tofu_* functions are now FROZEN and follow strict semantic versioning:
- MAJOR version: Breaking API changes only
- MINOR version: Backward-compatible additions
- PATCH version: Backward-compatible bug fixes
See API_STABILITY.md for full stability guarantees and deprecation policy.
✨ What's Included in v1.0.0
14 Core Operations
- Matrix operations: matmul, add, mul
- Activations: relu, softmax
- Normalization: layer_norm
- Shape ops: reshape, transpose
- Reductions: mean, sum
- Loss functions: MSE, cross-entropy
3 Optimizers
- SGD (stochastic gradient descent)
- SGD with momentum
- Adam
Quality Assurance ✅
- 70+ functions with comprehensive Doxygen documentation
- 13/13 validation tests passing
- 100% accuracy on CNN and ResNet examples
- Memory safe (AddressSanitizer verified, no leaks)
- CI/CD on Linux and macOS
- ESP32 cross-compilation support
⚠️ Known Limitations
These will be addressed in future releases:
- Performance benchmarks not yet established (planned for v1.1.0)
- Some broadcasting edge cases documented with TODOs
- Limited slice functionality (start < stop only)
🔄 Migration from v0.9.0
No breaking changes - v0.9.0 code works with v1.0.0 without modification.
🚀 What's Next
- v1.1.0: Performance benchmarks and optimizations
- v1.2.0+: Enhanced documentation (mdBook user guide and tutorials)
See CHANGELOG.md for detailed release notes.