Hugging Face's logo
Luigi's picture
291 5 82

Luigi PRO

luigi12345
Larry37's profile picture RustyTake-Off's profile picture saktibagchi's profile picture
·

AI & ML interests

None yet

Recent Activity

liked a Space about 24 hours ago
L-AI/Gemini-UI-Generator
updated a Space 1 day ago
luigi12345/python-worker
updated a Space 1 day ago
luigi12345/autoclient-worker
View all activity

Articles

Organizations

None yet

Posts 8

view post
Post
393
NEW LAUNCH! Apollo is a new family of open-source video language models by Meta, where 3B model outperforms most 7B models and 7B outperforms most 30B models 🧶

✨ the models come in 1.5B https://huggingface.co/Apollo-LMMs/Apollo-1_5B-t32, 3B https://huggingface.co/Apollo-LMMs/Apollo-3B-t32 and 7B https://huggingface.co/Apollo-LMMs/Apollo-7B-t32 with A2.0 license, based on Qwen1.5 & Qwen2
✨ the authors also release a benchmark dataset https://huggingface.co/spaces/Apollo-LMMs/ApolloBench

The paper has a lot of experiments (they trained 84 models!) about what makes the video LMs work ⏯️

Try the demo for best setup here https://huggingface.co/spaces/Apollo-LMMs/Apollo-3B
they evaluate sampling strategies, scaling laws for models and datasets, video representation and more!
> The authors find out that whatever design decision was applied to small models also scale properly when the model and dataset are scaled 📈 scaling dataset has diminishing returns for smaller models
> They evaluate frame sampling strategies, and find that FPS sampling is better than uniform sampling, and they find 8-32 tokens per frame optimal
> They also compare image encoders, they try a variation of models from shape optimized SigLIP to DINOv2
they find
google/siglip-so400m-patch14-384
to be most powerful 🔥
> they also compare freezing different parts of models, training all stages with some frozen parts give the best yield

They eventually release three models, where Apollo-3B outperforms most 7B models and Apollo 7B outperforms 30B models 🔥https://huggingface.co/HappyAIUser/Apollo-LMMs-Apollo-3B
view post
Post
700
CHATGPT.com o1-MINI FOR FREE? Is this a bug?? Wow, I just converted gpt-4o-mini to o1-mini for free! In ChatGPT.com ! Is this a bug? I used this prompt

use CoT logic extensively to output the longest and richest and most beautiful possible verison of this app, call it MelindaAI Autoimage and make it be able to create 7 up to images with different prompts *the promtp of the user with differnt word order except for the first words that are fixed

  <!DOCTYPE html> <html lang="en"> <head>   <meta charset="UTF-8">   <meta name="viewport" content="width=device-width, initial-scale=1.0" ...

Really got it fully working and behaving in the UI with the complete Logic Section of Thoughts. I mean no surprises as it was quite obvious it was just the same model with backend automated reprompting, but it is quite astonoshing to see it behaving just the same as if I had choosen o1-mini which is limit rated while this one is free and UNLIMITED! Thoughts?
View all posts

Collections 4