Miguel Alba

malba96

AI & ML interests

ML Engineer - Visual Language Models - Continual learning - Adversarial Robusness

Recent Activity

commented on an article about 2 months ago

SigLIP 2: A better multilingual vision language encoder

upvoted an article 5 months ago

SmolVLM Grows Smaller – Introducing the 250M & 500M Models!

updated a collection 5 months ago

Visual Fashion Model

View all activity

Organizations

commented on SigLIP 2: A better multilingual vision language encoder about 2 months ago

I think that example is very specific to multi-label classification. @praveenNathan if you want to continue the "pre-training" and use it for zero-shot classification (or create better vector representations for your specific domain) you can check open-clip implementation https://github.com/mlfoundations/open_clip ... they have the losses there for v1