search

NickyFot / EmoCLIP

Apache License 2.0
39 stars 3 forks source link

Will the pre-trained models be released? #2

Closed tanshuai0219 closed 5 months ago

pritamqu commented 5 months ago

+1

NickyFot commented 5 months ago

Thank you both for showing interest in our work! I have added a link to the weights used for the backbone in the downstream task.

SethGA commented 4 months ago

Hi there, I've downloaded the task weights and attempted to load the state dictionary into the model, but I encountered some missing keys related to the temporal transformer. A gist of a notebook can be found here Thanks for your attention :)

NickyFot commented 3 months ago

please set the number of layers to 2 when initialising the model (as described in the paper)