docarray VS transformers

I’ve heard of DocArray, a library for multimodal data in transit and Pytorch Lightning which is also a tool for multimodal data. These two sound like a promising solution, but I’m not sure how to use it with databases or cloud storage. Do I need to install any additional packages or dependencies?

I'm considering using these tools for a recommender system for analyzing text data like user reviews: DocArray and the EZ-MMLA Toolkit. Can anyone share their experience with the DocArray and EZ-MMLA Toolkit? I would love to hear about others' experiences before making a final decision.

I have been thinking about trying out DocArray and the EZ-MMLA Toolkit .. Has anyone had experience with these two projects?? Let me know what you think!

For some alternatives, I know there’s DocArray where you can handle text, image and audio data. is basically a toolbox for multimodal data and then there should be Haystack which is also let you build search systems and also has to do something with Transformers and LLMs.

DocArray to manipulate data and interact with the storage backend using document store.

It's already been two weeks since the last alpha release of DocArray v2. And since then a lot has happened — we've merged features we're really proud of, and we've cried tears of joy and misery trying to coerce Python into doing what we want. If you want to learn about interesting Python edge cases or follow the advancement of DocArray v2 development then you’ve come to the right place in this blog post!

The German Fashion12k dataset is available for free use by the Jina AI community. After logging into Jina AI Cloud, you can download it directly in DocArray format:

Jina AI’s DocArray library

You could try spaCy. This is the brains of the operation - an open-source NLP library for advanced NLP in Python. Another is DocArray - It's built on top of NumPy and Dask, and good for preprocessing, modeling, and analysis of text data.

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.

I am aware that using mixed precision reduces the stability of weight and there will be little consistency but don't expect it to be this much. I have attached the graph of evaluation metrics. If someone can give me some insight into this issue, that would be great.

In transformers, they tried really hard to have a single function or method to deal with both self and cross attention mechanisms, masking, positional and relative encodings, interpolation etc. While it allows a user to use the same function/method for any model, it has led to severe parameter bloat. Just compare the original implementation of llama by FAIR with the implementation by HF to get an idea.

pip install git+https://github.com/huggingface/transformers.git@main