clip-as-service VS transformers

I think training models with limited images can lead to overfitting, so I think you can try using a set of images with different poses. You might also want to try flipping or to help out the model so it gets to do different psoes. You might also want CLIP-as-a-service, but just know that pre-trained models isn't always be the best solution. My .02c

Imagine an AI-powered grep command, one that could process a film and find segments matching a text. With CLIP-as-service, you can do that. Here is the repo link, https://github.com/jina-ai/clip-as-service.

you could try Jina.ai's CLIP-as-a-Service: https://github.com/jina-ai/clip-as-service

Multisearch allows people to search with both text and images. With Open-Source project CLIP-as-service, you can use CLIP (a deep learning model by OpenAI) to do the same. Ask me if you have any questions?

Are you scraping these images or using any dataset? Do share the link, would love to play around with it. Would love to hear your feedback for clip-as-service (what I use in my example)?

Built CLIP-as-service, an open-source library to create embeddings of images and text using CLIP. These embeddings can be used to find the relevant images for any sentence. Note: you don't need to caption the images for this to work, and it is not just limited to objects in the image but an overall understanding built via CLIP neural network.

Built [CLIP-as-service](https://github.com/jina-ai/clip-as-service), an open-source library to create embeddings of images and text using CLIP.

Excited to share my new project CLIP-as-service, a high-scalability service for embedding images and text. It serve CLIP models with ONNX runtime and PyTorch JIT with 800QPS.

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.

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.