fix(causal): stabilize Colab causal intervention reproduction

[DaviBonetto]

Adds a Colab-friendly reproduction notebook that clones the repository automatically and runs the causal intervention sweep from the published branch.

Also fixes the hook shadowing bug in scripts/run_causal_intervention.py that caused the first all_layer sample to abort with UnboundLocalError inside the clamp path.

This keeps the reproduction runnable from a fresh Colab runtime without manual repository upload.

Summary by CodeRabbit

• Bug Fixes
  • Improved internal calculation handling in the causal intervention script to ensure reliable computation of temporal parameters without unintended side effects on intermediate values.

[coderabbitai]

Caution

Review failed

The pull request is closed.

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: 3a5af537-f73a-459b-9311-f0c499946585

📥 Commits

Reviewing files that changed from the base of the PR and between 5833a28 and 31edba9.

📒 Files selected for processing (1)

• scripts/run_causal_intervention.py

---

📝 Walkthrough

Walkthrough

Modified scripts/run_causal_intervention.py to introduce a new min_dt_effective variable computed from a dimensionally-constrained slice of min_dt_local, ensuring the original slice remains unmodified while downstream operations use the derived value for shaping, clamping, and logging.

Changes

Cohort / File(s)	Summary
Variable Extraction scripts/run_causal_intervention.py	Introduced min_dt_effective computed as min_dt_local[:dt_dim].to(...) and updated all downstream references (shaping, torch.maximum clamping, and diagnostic prints) to use the new variable instead of mutating min_dt_local directly.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~8 minutes

Poem

🐰 A slice once mutated, now sits pristine and still,
While min_dt_effective hops up the hill,
Correct dimensions flow where they're meant to go,
No side effects whisper—just a cleaner, safer flow! ✨

✨ Finishing Touches

• 📝 Generate docstrings (stacked PR)
• 📝 Generate docstrings (commit on current branch)

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch codex/fix/colab-causal-repro-notebook

📝 Coding Plan

• Generate coding plan for human review comments

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

Tip

CodeRabbit can approve the review once all CodeRabbit's comments are resolved.

Enable the reviews.request_changes_workflow setting to automatically approve the review once all CodeRabbit's comments are resolved.