search

lucidrains / bottleneck-transformer-pytorch

Implementation of Bottleneck Transformer in Pytorch
MIT License
667 stars 82 forks source link

model in README #12

Open red-lighting opened 3 years ago

red-lighting commented 3 years ago

Hello, The BoTNet you wrote in the README file is different from that in the article. Is it wrong or intentional? Thanks

nit-hmz commented 3 years ago

I found the same problem as "red-lighting" said: the README file is different? I think if use resnet50 as the backbone, so dim = 1024, *backbone[:-3].

SamREKKLES commented 2 years ago

I also meet the same problem, is it wrong in "*backbone[:5]"?

CropUpperY commented 1 month ago

I also meet the same problem, is it wrong in "*backbone[:5]"?

May I ask what to make of "*backbone[:5]", does it mean that the first 5 layers of resnet50 remain the same and all subsequent conv_x are replaced with BottleStack?

SamREKKLES commented 1 month ago

您的来信已经收到！我会尽快回复！

CropUpperY commented 1 month ago

I found the same problem as "red-lighting" said: the README file is different? I think if use resnet50 as the backbone, so dim = 1024, *backbone[:-3].

Does dim mean dimension, please? Why do you need to set *backbone[:-3] if you use resnet50, wouldn't you just use resnet50?