What are the different architectures available?

Part 1 (2020)
1

I know there are resnet18, resnet34, resnet50; but where can you find an exhaustive list of the architectures that can be used? I’ve tried looking in the docs but to no avail…

1 Like
2

You can find all of the built-in models/architectures in the fastai docs. Specifically, for vision models you can go to Vision → Models and look at the presets (there’s an Xresnet and a Unet page). Other tasks have their own model lists as well. They label them as models instead of architectures, which might be why you missed them.

There’s a lot more to try out (and with many more updated and sophisticated approaches) than the fastbook leads you to believe!

3 Likes
3

Thanks, I didn’t find this with a search, I guess I should use the menu more to find things. I still can’t find everything that you can use here though. For example, densenet201 can be used but I can’t find this in the fastai docs. Maybe it’s because these are from PyTorch and not fastai?

4

Think I’ve figured out how to find them all. You can go into the fastai github: pytorch models. This shows all the models fastai imports from pytorch. So I guess if you wanted one not on the list you could look here: torchvision.models — Torchvision 0.10.0 documentation (pytorch.org) and manually import them.

1 Like
5

Hello @nauzk,

Thanks a lot for update and quick reply mygiftcardsite, I have same question and i found lots of things here, Really appreciate for help.

1 Like