Changed Paper link from local PDF download to arXiv external link:
- URL: _static/downloads/TinyTorch-Paper.pdf → arxiv.org/abs/2601.19107
- Icon: ↓ (download) → ↗ (external link)
- Added target=_blank to open in new tab
The nbdev_export CLI was not reading settings.ini correctly in CI,
causing exports to silently fail. Using nbdev.export.nb_export()
Python API directly with explicit lib_path ensures exports work
reliably regardless of environment.
The nbdev_export command wasn't in PATH in CI. Using sys.executable
with -m nbdev.export ensures we use the same Python environment that's
running tito, which is more reliable across different environments.
Fixed in:
- tito/commands/dev/export.py (both _export_all_modules and _run_nbdev_export)
- tito/commands/module/workflow.py (_export_notebook)
The directory check was looking for tinytorch/tinytorch/__init__.py
when running from the tinytorch directory (as CI does). Fixed to check
for either repo root or tinytorch project directory structure.
- Default test_type: quick → all (run full validation by default)
- Default os: linux → all (test both platforms by default)
- Rename 'release' → 'user-journey' to match CLI --user-journey flag
- Fix 'standard' to correctly stop at stage 5 (was running user-journey)
- Reorder options to show defaults first in dropdown
- Change Validate (Dev) icon from 💯 to ✅ for Book and TinyTorch
- Change Update PDFs icon from 📄 to 📑 (bookmark tabs)
- Fix workflow_run references to match exact workflow names
This makes it easier to visually distinguish between Validate and
Publish operations in the GitHub Actions UI.
Fixed NameError in step_1_profile and step_5_accelerate by adding
Tensor as a parameter. These functions create sample tensors for
profiling/testing but the Tensor class is imported in main(), so
they need it passed as an argument.
Each challenge function now has visual documentation showing:
- The task input/output example
- The ideal attention weight matrix pattern
- Why the pattern is required for the task
Challenge 3 also explains why a fresh model is needed
(sequential training causes "catastrophic forgetting").
Break the monolithic main() into clean, documented functions:
- CONFIG dict for shared hyperparameters
- build_model() for creating fresh model/optimizer/loss
- challenge_1_reversal() - anti-diagonal attention patterns
- challenge_2_copying() - diagonal attention patterns
- challenge_3_mixed() - prefix-conditioned behavior (fresh model)
- print_final_results() - summary table and messages
This makes the code much easier for students to understand
and clearly shows why challenge 3 needs a fresh model.
The transformer was being trained sequentially on reversal then copying,
which caused it to "forget" reversal before the mixed task. Now we
reinitialize the model before challenge 3 so it learns both tasks
together with proper prefix conditioning.
- Skip "sync to profile" prompt when not in interactive terminal
(was hanging in CI waiting for input that never comes)
- Increase milestone timeout from 180s to 300s for heavy milestones
(CNN and Transformer training can take 2-3 minutes on CI)
Bug Fix:
- Export failures now return early in module complete, preventing
progress from being updated when export fails. This fixes the issue
where students could start Module 02 even though tensor.py was
never exported.
New --user-journey test:
- Simulates exact student flow: module start --no-jupyter → module complete
- Runs milestones at natural unlock checkpoints (not all at end)
- Milestone checkpoints: after modules 03, 08, 09, 13, 19
- Explicit step-by-step output for easy CI debugging
New --no-jupyter flag:
- Added to 'tito module start' for CI/testing
- Creates notebook but skips opening Jupyter
Ref: https://github.com/harvard-edge/cs249r_book/discussions/1147
- Name the problem: "algorithm-systems divide"
- Name the approach: "implementation-based systems pedagogy"
- Add concrete systems examples (O(N^2), Adam 3x memory)
- Include MLPerf-style benchmarking in milestones
- Strengthen citable terminology throughout
