why would I use a 200MB classifier when I can use a 40GB LLM named psiball-orpo-qdora-the-xplora-70B-int4-swiffer-sweeper-slerp-v0.02-(Taylor's version)

if you want a weekend short watch this talk goes through foundation models in computer vision and how to use/adapt/optimize them 🏄🏻‍♀️

Mantis’s technical report has now been released!

🔥Summary of updates:
🚀Mantis-8B-SigLIP-LLaMA-3 archives the best open-source performance on 5 multi-image leaderboards (NLVR2, Q-Bench, BLINK, MVBench, Mantis-Eval)
⚙️Mantis-8B-Fuyu is released as a support of super-resolution!…

Yapay zeka - makine öğrenmesi tarafında halihazırda ücretsiz, açık kaynak içerik paylaşmayı seven arkadaşlarımla birlikte (merve, Tarık Karakaş, Kadir Nar, Mert Cobanov) bunu bir community-driven şekle sokmaya karar verdik.

twitter.com/i/communities/…

Cutting-edge Computer Vision with Transformers youtube.com/watch?v=BnM-S5…

Another quantization method dropped in Hugging Face transformers library ! Half Quadratic Quantization 🔥

HQQ implements on-the-fly quantization via fast robust optimization. It doesn’t require calibration data and can be used to quantize any model, up to 1-bit precision !

Thanks merve and Linoy Tsaban🎗️ for trying our DoRA ICML Conference for diffusion models, generating better details aligned with text prompts!!

Feel free to try the official code: github.com/NVlabs/DoRA
Stay tuned for more details!

#ICML2024 #icml #peft #lora #AI #ML