AngelBottomless

No, I mean it would be better to have random information, and unique rng that is only used for dataset shuffling. In this way, we can reproduce training in same...

@enn-nafnlaus Can you specify the setups which were broken? At least I see Linear to be broken due to my mistake (its urgent, sorry, there's existing PR mentioned above) but...

Since there are results from[ mish activation and xavier normal](https://arca.live/b/hypernetworks/61555752?p=1) I conclude that there are no missing functions except missing Linear. There are far more results that proves [swish is...

@gzalwa201 The only change I catched related to linear is [weird 0.005 from nowhere ](https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/3698/commits/029d7c75436558f1e884bb127caed73caaecb83a) , which should be fixed with returning of linear functions. And note that even with...

https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/3717 Currently PR is open, it will be merged soon. Sorry for the inconvenience.

It is safety feature to prevent continuing failed attempt of training : I'll note that ML is not perfect, you'll run into some *many* failuares and some successes. There're no...

You should never use `eval` in general, especially in release codes anywhere. Rather consider `ast.literal_eval` for safety or `re` to save and parse.

by changing `startswith` as dict type check as literal eval result, and using `.update(another dict)`, this should be better. Currently raises exception when unexpected or invalid comments are in exif.

Checked safety in low-bandwidth-network in gcp. I can't find related issues with pip read time out, did you have local error with this? When time out error happens, it more...

@AUTOMATIC1111 Yes, AdamW (and Momentum based optimizers) uses adaptive learning rate, which is estimated from its momentum. If we try to start from zero, AdamW will use given pure learning...