github-starred/cs249r_book
2
0
Fork 0
mirror of https://github.com/harvard-edge/cs249r_book.git synced 2026-05-01 18:19:18 -05:00
Code Issues 77 Packages Projects Releases 24 Wiki Activity

fix: complete module renumbering across entire codebase

Browse Source 
Updated all references to reflect new module order:
- Module 05: DataLoader (was 08)
- Module 06: Autograd (was 05)
- Module 07: Optimizers (was 06)
- Module 08: Training (was 07)

Changes include:
- paper/paper.tex: 20+ references, tier descriptions, milestones
- src/: Export commands, dependency diagrams, docstrings
- tests/: Dependency chains, integration tests, README
- tito/: export_utils.py path mappings
- tinytorch/: Auto-generated package file headers

Foundation Tier is now Modules 01-08
Architecture Tier is now Modules 09-13
This commit is contained in:
Vijay Janapa Reddi
2025-12-19 17:43:41 -05:00
parent b3eb1aa90a
commit d203fba8b8
69 changed files with 292 additions and 293 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
tinytorch/tests/regression/test_gradient_flow_fixes.py
View File

@@ -78,9 +78,9 @@ def test_regression_subtraction_has_backward():
"""
Regression test for Issue #3: Subtraction had no backward pass.
Bug: Tensor.__sub__ not patched by Module 05, no gradient flow.
Fix: Added SubBackward class and patched __sub__ in Module 05.
Commit: Module 05 fixes
Bug: Tensor.__sub__ not patched by Module 06, no gradient flow.
Fix: Added SubBackward class and patched __sub__ in Module 06.
Commit: Module 06 fixes
"""
print("Testing regression: subtraction backward...")
@@ -105,9 +105,9 @@ def test_regression_division_has_backward():
"""
Regression test for Issue #4: Division had no backward pass.
Bug: Tensor.__truediv__ not patched by Module 05, no gradient flow.
Fix: Added DivBackward class and patched __truediv__ in Module 05.
Commit: Module 05 fixes
Bug: Tensor.__truediv__ not patched by Module 06, no gradient flow.
Fix: Added DivBackward class and patched __truediv__ in Module 06.
Commit: Module 06 fixes
"""
print("Testing regression: division backward...")
@@ -213,9 +213,9 @@ def test_regression_transpose_has_backward():
"""
Regression test for Issue #8: Transpose had no backward pass.
Bug: Tensor.transpose() not patched by Module 05, no gradient flow.
Fix: Added TransposeBackward class and patched transpose in Module 05.
Commit: Module 05 fixes (TransposeBackward)
Bug: Tensor.transpose() not patched by Module 06, no gradient flow.
Fix: Added TransposeBackward class and patched transpose in Module 06.
Commit: Module 06 fixes (TransposeBackward)
"""
print("Testing regression: transpose backward...")
@@ -243,8 +243,8 @@ def test_regression_matmul_backward_uses_matmul():
Regression test for Issue #9: MatmulBackward used np.dot for gradients.
Bug: MatmulBackward used np.dot which doesn't handle batched 3D+ tensors.
Fix: Changed to np.matmul and np.swapaxes in Module 05.
Commit: Module 05 fixes (MatmulBackward batched)
Fix: Changed to np.matmul and np.swapaxes in Module 06.
Commit: Module 06 fixes (MatmulBackward batched)
"""
print("Testing regression: MatmulBackward batched operations...")