Level up¶

Learn enough Lightning to match the level of expertise required by your research or job.

Join our community

Basic skills¶

Learn the basics of model development with Lightning. Researchers and machine learning engineers should start here.

Level 1: Train a model

Learn the basics of training a model.

basic

Level 2: Add a validation and test set

Add validation and test sets to avoid over/underfitting.

basic

Level 3: Use pretrained models

Learn how to use pretrained models with Lightning

basic

Level 4: Enable script parameters

Add parameters to your script so you can run from the commandline.

basic

Level 5: Understand and visualize your model

Remove bottlenecks and visualize your model

basic

Level 6: Predict with your model

Use your model for predictions.

basic

Intermediate skills¶

Learn to scale up your models and enable collaborative model development at academic or industry research labs.

Level 7: Interactive cloud development

Learn how to access GPUs and TPUs on the cloud.

intermediate

Level 8: Train in the background on the cloud

Learn how to run models on the cloud in the background.

intermediate

Level 9: Modularize your projects

Create DataModules to enable dataset reusability.

intermediate

Level 10: Understand your model

Use advanced visuals to find the best performing model.

intermediate

Level 11: Explore SOTA scaling techniques

Explore SOTA techniques to help convergence, stability and scalability.

intermediate

Level 12: Deploy your models

Learn how to deploy your models with optimizations like ONNX and torchscript.

intermediate

Level 13: Optimize training speed

Use advanced profilers to mixed precision to train bigger models, faster.

intermediate

Level 14: Run on on-prem clusters

Run on a custom on-prem cluster or SLURM cluster.

intermediate

Advanced skills¶

Configure all aspects of Lightning for advanced usecases.

Level 15: Customize configs to run in production

Enable composable YAMLs

advanced

Level 16: Customize the trainer

Inject custom code into the trainer and modify the progress bar.

advanced

Level 17: Own the training loop

Learn all the ways of owning your raw PyTorch loops with Lighting.

advanced

Level 18: Enable advanced checkpointing

Enable composable or cloud based checkpoints.

advanced

Level 19: Explore IPUs

Explore Intelligence Processing Unit (IPU) for model scaling.

advanced

Level 20: Explore HPUs

Explore Havana Gaudi Processing Unit (HPU) for model scaling.

advanced

Level 21: Master TPUs

Master TPUs and run on cloud TPUs.

advanced

Level 22: Reach 1 trillion parameters on GPUs

Scale to 1 trillion params on GPUs.

advanced

Expert skills¶

Customize and extend Lightning for things like custom hardware or distributed strategies.

Level 23: Extend the Lightning CLI

Extend the functionality of the Lightning CLI.

expert

Level 24: Integrate a custom cluster

Integrate a custom cluster into Lightning.

expert

Level 25: Explore fault-tolerance in-depth

Understand the details of fault-tolerance.

expert

Level 26: Make your own profiler

Make your own profiler.

expert

Level 27: Add a new accelerator or Strategy

Integrate a new accelerator or distributed strategy.

expert