Where are log files/training arguments that were used to achieve each model here?

[Coderx7]

Hi, if I remember correctly, we could once see how for example mobilenet model was trained and which set of arguments/hyperparameters were used. I cant seem to find it anywhere, eventhough the readme says something should be in the root of the repository.
How can I know for example how a specific model was trained so that I can replicate the result on my end as well?
Thanks a lot in advance

[Coderx7]

@rwightman any update on this would be greatly appreciated!

[rwightman]

@Coderx7 generally they don't exist, I have them for quite a few models, but maintaining and updating (as args and library evolves) / answering questions about them is too much work for me so I haven't been releasing them.

That said, others have asked about mobilenet hparams and you can find them if you sarch issues / discussions...

[natanielruiz]

I understand the problem of releasing + updating - but is there not the possiblity of just releasing a text file as an archival piece? That would help many people do quick comparisons without re-training the base model.