Call for High Quality Lightning Lectures - Community Sprint

[tchaton]

🚀 Feature

Dear community,

As PyTorch Lightning mature, we believe it is important for the Lightning Team and its community to improve the Lightning onboarding process.

In that regards, the PyTorch Lightning Team (@Borda) built a https://github.com/PyTorchLightning/lightning-tutorials repo containing tuto-books (notebooks as python script). The tuto-books are converted into notebooks, fully tested on GPUs and injected directly within the Lightning documentation.
Learn more with this blogpost: https://devblog.pytorchlightning.ai/publishing-lightning-tutorials-cbea3eaa4b2c

The tutorials can be found there on the main page of the documentation: https://pytorch-lightning.readthedocs.io/en/latest/

@phlippe, PHD from the University of Amsterdam contributed their Deep Learning Course from 2020 and it is now rendered within Lightning documentation (released officially with v1.5). Great thanks from the entire community :)

However, the Lightning Team doesn't want to stop there and we would like to encourage the community to contribute more high quality course and collaborate closely with us to implement best practices while doing so.

We plan to organize a tutorial sprint to create high quality tutorial / course and would like the community to share ideas and collaborate with the Lighting Team in this project. We will soon share a Google Form for the community to share their ideas and possibly to contribute.

Here are the requirements for a tutorial to be accepted:

• The Tutorial should present a field of AI in simple terms, not a single method. The UVA Course SSL or Meta-Learning notebooks are a great example of such works.

• The Tutorial should contain high quality animated visuals, be engaging, well written and simple to follow. Assume new comers.

• The tutorial should come with a short description, thumbnail.

• The Tutorial should contain a YouTube video going through the notebook + attendees asking questions (Lightning Team and community). We might live stream the event and use the recording as its explanation ;)

Note: The Lightning Team Editors can help there with visuals, thumbnail

Here are some high quality notebooks we are planning to work on:
(This list isn't exhaustive and we would like the community to propose ideas based on their skills and interests).

• Lightning introduction for engineers
• Lightning introduction to distributed training and challenges. Ultimate guide.
• Lightning best practices
• Lightning introduction for researchers
• Lightning and Video (PyTorch Video)
• Lightning and PointCloud
• Lightning and 3D Rendering (PyTorch3D)
• Lightning and Quantum (Pennylane)
• Lightning and Probabilistic Models (scvi-tools)
• Lightning and Forecasting (PyTorch Forecasting)
• Lightning on Kaggle (Flash)
• Lightning and 3D Images (Monai)
• Lightning and Temporal Graphs (PyTorch Geometric / PyTorch Geometric Temporal)
• Lightning and GANS
• Lightning and Style Transfer
• Lightning and Image Synthesis (https://github.com/CompVis/taming-transformers)
• Lightning and Reinforcement Learning (Single / Multiple Agents)
• Lightning and Online Learning
• Lightning and Sequential Learning
• Lightning and ODE
• Lightning and Federated Learning
• Lightning and Multi Tasks
• Lightning and HPO (Optuna)
• Lightning and Precision Scheduling
• Lightning and Arts Generation !

# Applications related

• PyTorch Profiler + ORTModule + DeepSpeed Sparse + OnnxRuntime.js + Triton (Exploration for optimized application building)

Bonus: Notebook translation would be a massive plus and would help to democratize AI in the entire world :)

• Chinese
• German
• Czech
• Hebrew
• French
• Russian
• Spanish
• Japanase
• Hindi

If you are interested, please add a comment below and assign your name to the proposal(s), your level, share about your availabilities and your confidence into your contribution being impactful from 1 to 5 :)

Please, apply to this Google Form

---

If you enjoy Lightning, check out our other projects! ⚡

• Metrics: Machine learning metrics for distributed, scalable PyTorch applications.

• Flash: The fastest way to get a Lightning baseline! A collection of tasks for fast prototyping, baselining, finetuning and solving problems with deep learning

• Bolts: Pretrained SOTA Deep Learning models, callbacks and more for research and production with PyTorch Lightning and PyTorch

• Lightning Transformers: Flexible interface for high performance research using SOTA Transformers leveraging Pytorch Lightning, Transformers, and Hydra.

[Borda]

just a quick question, so we expect some overlap with existing tutorials and lectures from UvA or shall be rather very unique?

Bonus: Notebook translation would be a massive plus and would help to democratize AI in the entire world :)

get think about some automatic translation which we can do in build time...

[tchaton]

Dear @Borda,

I believe it would be more fruitful to prevent overlap with existing tutorials and lectures from UvA.

The goal is to create Lightning Lectures for user to learn how to utilise Lightning for their own research but more importantly, be able venture and learn more broadly about Lightning usage in different fields from a centralised place, so they can utilise those ideas and concepts for their own use case and ultimately become more proficient.

Get automatic translation would be great, but I am a bit afraid of mis-translation. I think it is fine if reviewed by a native speaker before being publicly released.

Best,
T.C

[akihironitta]

Get automatic translation would be great, but I am a bit afraid of mis-translation. I think it is fine if reviewed by a native speaker before being publicly released.

I agree on this since the maintenance cost wouldn't be too large as notebooks don't change much over time while the docs can be outdated pretty quickly.