tfjs VS transformers

TensorFlow.js

Many people seem to be unaware of tensorflow.js, an official JS implementation of TF

https://github.com/tensorflow/tfjs

I'd love to see PyTorch in JS, but I think unless you get it running on the GPU it won't be able to do much.

TensorflowJS GitHub Repository

Also, Tensorflow.js WebGPU backend has been in the works for quite some time: https://github.com/tensorflow/tfjs/tree/master/tfjs-backend-...

It's a pity that tfjs never truly developed any decent ops. E.g. you need lgamma to implement the cap for zero-inflated poisson regression and tfjs simply doesn't have that: https://github.com/tensorflow/tfjs/issues/2011

People have been doing it for long with WebGL, see eg https://github.com/tensorflow/tfjs and https://cloudblogs.microsoft.com/opensource/2021/09/02/onnx-...

thanks, no not unity, going to show it as a demo with threejs + tensorflow on the web. I found a github request to add face orientation https://github.com/tensorflow/tfjs/issues/3835 looks like they assigned someone to add it but doesn't look like its available yet, but there's some posts about the math I can use to get rotations based on some of the landmarks

There already exist a wide variety of neural network inference engines that run in web browsers (e.g. TensorFlow.js and, my personal favorite for use with PyTorch models, ONNX Runtime Web), but pre- and post-processing has always required imperative manipulations on flat buffers rather than a clean ndarray interface.

Full docs and code: https://github.com/tensorflow/tfjs/tree/master/e2e/benchmarks/local-benchmark

transformers is a package by Hugging Face, that helps you interact with models on HF Hub (GitHub)

Is t5x an encoder/decoder architecture?

Some more general options.

The Flax ecosystem

https://github.com/google/flax?tab=readme-ov-file

or dm-haiku

https://github.com/google-deepmind/dm-haiku

were some of the best developed communities in the Jax AI field

Perhaps the “trax” repo? https://github.com/google/trax

Some HF examples https://github.com/huggingface/transformers/tree/main/exampl...

Sadly it seems much of the work is proprietary these days, but one example could be Grok-1, if you customize the details. https://github.com/xai-org/grok-1/blob/main/run.py

The HuggingFace transformers library already has support for a similar method called prompt lookup decoding that uses the existing context to generate an ngram model: https://github.com/huggingface/transformers/issues/27722

I don't think it would be that hard to switch it out for a pretrained ngram model.

Most of this tutorial is based on Hugging Face course about Transformers and on Niels Rogge's Transformers tutorials: make sure to check their work and give them a star on GitHub, if you please ❤️

* Superconvergence + LR range finder + Fast AI's Ranger21 optimizer was the goto optimizer for CNNs, and worked fabulously well, but on transformers, the learning rate range finder sadi 1e-3 was the best, whilst 1e-5 was better. However, the 1 cycle learning rate stuck. https://github.com/huggingface/transformers/issues/16013

Thanks! :) I'm pushing them into transformers, pytorch-gemma and collabing with the Gemma team to resolve all the issues :)

The RoPE fix should already be in transformers 4.38.2: https://github.com/huggingface/transformers/pull/29285

My main PR for transformers which fixes most of the issues (some still left): https://github.com/huggingface/transformers/pull/29402

If you want to tinker with the architecture Hugging Face has a FOSS implementation in transformers: https://github.com/huggingface/transformers/blob/main/src/tr...

If you want to reproduce the training pipeline, you couldn't do that even if you wanted to because you don't have access to thousands of A100s.