mirror of
https://github.com/MLSysBook/TinyTorch.git
synced 2026-03-11 21:43:34 -05:00
97e0563614
- Implement tito benchmark baseline and capstone commands - Add SPEC-style normalization for baseline benchmarks - Implement tito community join, update, leave, stats, profile commands - Use project-local storage (.tinytorch/) for user data - Add privacy-by-design with explicit consent prompts - Update site documentation for community and benchmark features - Add Marimo integration for online notebooks - Clean up redundant milestone setup exploration docs - Finalize baseline design: fast setup validation (~1 second) with normalized results
3.6 KiB
3.6 KiB
Cloud Notebook Options for TinyTorch
Current Setup
Currently Configured:
- ✅ MyBinder (
https://mybinder.org) - Free, open-source, works well - ✅ Google Colab (
https://colab.research.google.com) - Free, popular, GPU access
Available Options
1. MyBinder (Current) ✅
Pros:
- Free and open-source
- No account required
- Works directly from GitHub
- Good for educational use
- Already configured and working
Cons:
- Can be slow to start (2-5 minutes)
- Limited resources (CPU, memory)
- No GPU access
- Sessions timeout after inactivity
Best For: Educational use, quick demos, zero-setup access
2. Google Colab (Current) ✅
Pros:
- Free tier available
- GPU access (free tier: T4 GPU)
- Fast startup
- Popular and familiar to students
- Good integration with Google Drive
Cons:
- Requires Google account
- Free tier has usage limits
- Sessions disconnect after inactivity
- Can be slow during peak times
Best For: Students who need GPU, familiar Google ecosystem
3. Deepnote (Not Currently Configured)
Pros:
- Modern, polished interface
- Real-time collaboration
- Good for team projects
- Free tier available
- Better than Colab for some use cases
Cons:
- Less well-known than Colab
- Requires account
- Free tier limitations
Best For: Team collaboration, professional workflows
How to Add:
# In site/_config.yml
launch_buttons:
deepnote_url: "https://deepnote.com"
4. JupyterHub (For Institutions)
Pros:
- Self-hosted control
- Institutional integration
- Can provide GPUs
- Scalable
Cons:
- Requires server infrastructure
- Setup complexity
- Maintenance overhead
Best For: Universities, institutions with IT support
5. Kaggle Notebooks
Pros:
- Free GPU access
- Popular ML community
- Good for competitions
Cons:
- Less flexible than Colab
- More focused on competitions
Best For: ML competitions, Kaggle users
Recommendation for TinyTorch
Current Setup is Good ✅
MyBinder + Colab covers most use cases:
- MyBinder: Zero-setup, no account needed, perfect for quick access
- Colab: GPU access when needed, familiar to students
Optional Addition: Deepnote
If you want to add Deepnote for better collaboration:
-
Add to config:
launch_buttons: binderhub_url: "https://mybinder.org" colab_url: "https://colab.research.google.com" deepnote_url: "https://deepnote.com" # Add this -
Benefits:
- Better collaboration features
- More modern interface
- Good for team projects
-
Considerations:
- Adds another option (might be confusing)
- Students need to create account
- Current setup already works well
What About "Mariomi"?
I couldn't find a tool called "Mariomi" related to notebooks. You might be thinking of:
- MyST (MyST Markdown) - Already used by Jupyter Book (for documentation)
- Miro - Collaboration whiteboard (not for notebooks)
- Deepnote - Modern notebook platform (see above)
My Recommendation
Keep current setup (MyBinder + Colab) because:
- ✅ Already working
- ✅ Covers all use cases
- ✅ No additional complexity
- ✅ Students familiar with Colab
- ✅ MyBinder perfect for zero-setup access
Optional: Add Deepnote if you want better collaboration features, but it's not necessary.
Testing Current Setup
To verify launch buttons work:
- Build site:
cd site && make html - Check notebook pages have launch buttons
- Test Binder: Click "Launch Binder" → Should open MyBinder
- Test Colab: Click "Launch Colab" → Should open in Colab