DLYuanGod/TinyGPT-V

TinyGPT-V: Efficient Multimodal Large Language Model via Small Backbones

3 mentions on Hacker News by 3 people: 1 story, 2 comments · first 2024-01-03 · latest 2024-01-04

Mentions per year 3 mentions · busiest year 2024 (3)

20082026

All Stories Comments Newest Top Oldest

…the dependencies of TinyGPT-V you'll see that it does not depend on tinygrad but rather torch... https://github.com/DLYuanGod/TinyGPT-V/blob/main/environment...
comment by roflmaostc · 2024-01-04 · in TinyGPT-V: Efficient Multimodal Large Language Model via Small Backbones
Being multimodal doesn’t seem to require much of a size penalty: https://github.com/dlyuangod/TinyGPT-V Even so, Google treats the Gemini Pro Vision model as a separate model from Gemini Pro, so it could have separate parameters that are…
comment by coder543 · 2024-01-03 · in Gemini in Reasoning: Unveiling Commonsense in Multimodal Large Language Models