model_name=="inception"

[Harry-KIT]

Hi @marvinbuss , good job. I have one question. Does the below model_name always have to be inception? I do not understand why it is like that?

is_inception=(model_name=="inception")

[marvinbuss]

Hi @Harry-KIT,
Thanks for your question!
You can use any of the models as baseline and you can also extend this with newer models that are available:

The inception model has a different architecture and needs to be treated slightly different compared to other models like ResNet. Inception uses auxiliary classifiers, which need to be handled in the output of the model. This is why there is a special check whether an inception based model is used.

Today, I would probably code this a bit nicer. Happy to accept any PRs from your side to improve the notebook.
Does that answer your question?

[Harry-KIT]

I got it
Thank you @marvinbuss ^^.

[marvinbuss]

Happy to help! You are welcome!

[Harry-KIT]

Hi @marvinbuss
Do you have a future plan to make multi-gpus training parse on this repo?

[marvinbuss]

Hi @Harry-KIT,
I had no plans to add this at this point in time. Would you be willing to contribute and add this?

[Harry-KIT]

ezgif.com-gif-maker.mp4

https://github.com/Harry-KIT/Image-Classification-on-small-datasets-in-Pytorch
Hi @marvinbuss. I created similar repo based on your github repo and some other useful sources. it includes multi-gpu option as well. Thank you for that you inspired me to learn many things from your repo.

[marvinbuss]

Hi @Harry-KIT, Sorry for my very late response here. This is awesome. Great work and thanks for sharing! 👍