-
|
Hi, Thank you for the great work. I understand pretrained model weights either come from the original source, or from models that you trained within the timm framework. Is there documentation for which weights come from where? It seems that models trained within the timm framework may use much heavier augmentation. Controlling for this is important for the sake of my study. For example, it seems like the ResNet50 was trained within timm, but other ResNets (18,34) may have been ported from the original source (this is a hunch based on the results I am getting in my study) Thank you |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment
-
|
Ah, I see in each individual {model}.py file, there are links to sources of weights (e.g. 'https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/resnet.py') Closing this question for now, though it would be nice if links to config files/description of training procedure for provided model weights were included as an enhancement. thx! |
Beta Was this translation helpful? Give feedback.
Ah, I see in each individual {model}.py file, there are links to sources of weights (e.g. 'https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/resnet.py')
Closing this question for now, though it would be nice if links to config files/description of training procedure for provided model weights were included as an enhancement.
thx!