unifyai / demos Goto Github PK

Currently the list of notebooks uses the paths (which is the intended order), but the navigation at the bottom of the page uses the names (which is obv a different order)

This issue lists all the examples that we should work on, if you have suggestions about new examples you can post them here.

If you want to work on any of these examples, please create an issue with the same name as the list-item bold name and edit this comment to replace the name with the issue link so that everybody can know who is working on what 😄 .

When creating a new notebook, please use the name of the list item in camel case (i.e. First Example with Ivy -> first_example_with_ivy.ipynb).

A few points:

• There will be tags classifying notebooks in related to "Ivy as a framework" and related to "Ivy as a transpiler"
• Notebooks with 🚨 should be prioritized!
• Notebooks in learn the basics should be extremely short and to the point, tutorials should be mid-length, with examples and demos probably being more lengthy and involved

• 🚨 Quickstart (framework)(transpiler)

Learn the basics

• 🚨 Write framework-agnostic code (framework)
• 🚨 Unify code (transpiler)
• 🚨 Compile code (framework)(transpiler)
• 🚨 Transpile code (transpiler)
• 🚨 Lazy vs Eager (framework)(transpiler)
• 🚨 How to use decorators (framework)(transpiler)
• 🚨 Transpile any library (transpiler)
• 🚨 Transpile any model (transpiler)
• 🚨 Write a module using Ivy (framework)

Tutorials

• 🚨 Developing a convolutional network using Ivy (framework)
• Developing a transformer using Ivy (framework)
• 🚨 Transpiling a torch model to build on top of it (transpiler)
• #15
• 🚨 Transpiling a haiku model to build on top of it (transpiler)

Examples and demos

• 🚨 #6
• 🚨 (?) Developing DeepMind's PerceiverIO using Ivy (framework) (?)
• Deploying a PyTorch model using TFLite (transpiler)
• Running a timm model in JAX using TPUs (transpiler)
• Benchmarking performance - Benchmark of the performance of a complex model after transpiling it to different backends (transpiler)
• Using fvcore to measure FLOPs in a keras model - this example should explore this kind of tools and showcase how transpiling models can help compare them in a more fair and easy way (e.g. compare various models in different frameworks calculating the FLOPs with the same tool) (transpiler)
• Transpiling a model from Hugging Face - add something else here (transpiler)

To be categorized (probably in the basics):

• Debugging graphs (framework)(transpiler)
• Exploring the graph (show_graph) (framework)(transpiler)

Pretty much self-explanatory, an open space to discuss, propose, or suggest new examples or changes to the existing ones!

Task is to improve the existing Ivy Deepmind PerceiverIO notebook. Add more explanations, examples, and clarifications. Please assign youself, if you are interested.

https://github.com/unifyai/demos/blob/main/deepmind_perceiverio.ipynb