search

Luodian / Otter

🦦 Otter, a multi-modal model based on OpenFlamingo (open-sourced version of DeepMind's Flamingo), trained on MIMIC-IT and showcasing improved instruction-following and in-context learning ability.
https://otter-ntu.github.io/
MIT License
3.52k stars 241 forks source link

pre-trained weights #302

Closed hxan closed 7 months ago

hxan commented 8 months ago

Hello, authors, among the series of pre-trained weights of huggingface, are the weights of Llama all the same?

Luodian commented 8 months ago

Yes it's same. But it may differ from a pure llama weights since openflamingo would add cross gates attention layers into llm decoder layers.

hxan commented 8 months ago

Hello, author, would you like to ask, does it mean that I only need to download one of the pre-trained weights of about 30GB, and then change the pre-trained model by downloading different final_weights.pt? If so this will help me save a lot of download time