Skip to content

Tutorials/Accelerated Python: Notebook improvements and consistency fixes#151

Open
brycelelbach wants to merge 17 commits intomainfrom
fix/accelerated-python-notebook-improvements
Open

Tutorials/Accelerated Python: Notebook improvements and consistency fixes#151
brycelelbach wants to merge 17 commits intomainfrom
fix/accelerated-python-notebook-improvements

Conversation

@brycelelbach
Copy link
Collaborator

Summary

  • Benchmark reliability: Add n_warmup to all cupyx.profiler.benchmark calls and adjust matrix sizes for consistent GPU speedups.
  • Notebook consistency: Audit and sync all regular notebooks with their solution counterparts (notebooks 01-06), fixing typos, wording mismatches, and formatting inconsistencies.
  • Content improvements: Clarify memory/device explanations (notebook 03), expand View vs Copy coverage (notebook 01), use binary units for byte-to-GB conversions, label eigenvalue prints (notebook 05), and write checkpoints to /tmp for better performance on network filesystems (notebooks 05-06).
  • Libraries & kernels fixes: Fix CLI arg bugs, sync profiling text, correct typos, and re-canonicalize formatting across cuDF, cuda-cccl, copy kernel, and black-and-white kernel notebooks.
  • Add analytic solution assertions and tune convergence check frequency in power iteration notebooks (05, 06).
  • Documentation: Add pre-commit hooks documentation to CONTRIBUTING.md.

…atrix sizes for reliable speedups.

Ensure all cupyx.profiler.benchmark calls specify n_warmup=1 to avoid
cold-start measurement noise. Adjust SVD benchmark matrix shape from
(1000, 1000) to (3000, 1000) so GPU speedups are shown reliably in the
production environment.

Made-with: Cursor
…in notebook 03.

Simplify sequential operations intro, add Copy vs View pitfall note,
clarify that GPU copies stay on device, and clean up device management text.

Made-with: Cursor
…iew vs Copy explanation in notebook 01.

Made-with: Cursor
…nd solutions.

Audit and sync all regular notebooks with their solution counterparts:
- 02: Sync "Then we read" wording, add JPEG size comparison to solution.
- 03: Use self-contained benchmark cells with shape variables (matching solution).
- 04: Fix "signular" typo, sync tiling factor to (3,3), fix "differ sizes"
  typo, sync wording and quote style in solution.
- 05: Sync estimate_host comments/formatting, fix docstring steps, preserve
  "Think About It" question in solution, correct GPU memory description.
- 06: Fix pseudo-code typo in solution, sync Perfetto display title.

Made-with: Cursor
…rnels notebooks.

- 20 (cuDF): Sync import order, add %%time to Task 2 placeholder.
- 23 (cuda-cccl): Fix "excercise" typos, sync capstone print statement
  and remove stale TODO markers from solution iterator lines.
- 40 (copy kernel): Fix CLI arg bug (output -> check), sync profiling
  description and benchmarking text with solution.
- 43 (black and white): Fix hint to say 256 (matching solution's
  (16,16) block size), add Colab warning to solution.
- Re-canonicalize notebook formatting.

Made-with: Cursor
…05 and 06.

Fixes performance issues when the Git repository is on a network or slow
filesystem by writing checkpoint files to /tmp instead of the working directory.
… SOLUTION notebook outputs.

Add validation and auto-fix for execute_result and display_data outputs
missing required nbformat fields, which caused test-links CI to fail
during nbconvert.
@NVIDIA NVIDIA deleted a comment from github-actions bot Mar 9, 2026
Copy link
Collaborator

@nv-kriehl nv-kriehl left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Please use function names with parentheses for function names as well as function calls. .foo -> .foo()
Leave parentheses off for attribute names

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants