look into ONNX runtime - Githubissues

epfml / disco

DISCO is a code-free and installation-free browser platform that allows any non-technical user to collaboratively train machine learning models without sharing any private data.

https://discolab.ai

Apache License 2.0

154 stars 26 forks source link

look into ONNX runtime #650

Open tharvik opened 8 months ago

tharvik commented 8 months ago

ONNX has a JavaScript API, it seems to be the new framework on the bloc.

currently, we are using TensorFlowJS everywhere. to support more runtime, we would need to first abstract from it.

some of our types are typedef of tf's ones: Dataset, Weights
aggregation, validator, data_loader, preprocessing are all using tf types

JulienVig commented 8 months ago

MNIST web training with ONNX

martinjaggi commented 8 months ago

roadmap (not very LLM specific but good)

transformers.js uses it, and would easily give us tokenizers and model converters to/from HF which is crucial

bhack commented 2 months ago

https://opensource.microsoft.com/blog/2024/02/06/on-device-training-training-a-model-in-browser/

tharvik commented 4 days ago

bad news, upstream removed training from their web package.

JulienVig commented 4 days ago

😭

bhack commented 4 days ago

Probably one way to gain momentum is to push for training in the new W3C forming standards: https://github.com/webmachinelearning/webnn/blob/main/explainer.md?plain=1#L187