Sung Kim

sungkim.bsky.social

14 February 2025 12:56

Want to train a multimodal models with 100 billion 🚀 unique examples? Well, Google did that. Google's Scaling Pre-training to One Hundred Billion Data for Vision Language Models Paper Link: arxiv.org/abs/2502.07617

Sung Kim

sungkim.bsky.social

14 February 2025 10:01

Xiaomi's TransMLA: Multi-head Latent Attention Is All You Need MLA tackles the challenge of communication bottlenecks on current hardware by using low-rank matrices in the key-value (KV) layers, thereby allowing compressed latent KV states to be cached.

Isabelle Roussy

izylarousse.bsky.social

10 February 2025 15:08

Bon lundi éverébodé ! 🤭😃

Léo

leoauteur.bsky.social

10 February 2025 15:06

Magnifique comme tout… 📷 John Vargas

Mickaël Correia

mickacorreia.bsky.social

10 February 2025 15:06

Dingue, on ne s'y attendait tellement pas 🙃.