Ross Wightman

@Zasder3 I moved timm models to own file in recent PR, I can take a crack at moving the text transformer to a submodule in CLIP, and adding the checkpoint...

As further comment to #88 and detail on my previous comment There is some groundwork mixed up in my grad cache PR https://github.com/rwightman/open_clip/blob/grad_cache/src/open_clip/model.py for impl gradient caching, it was easier...

@arampacha @iejMac that sounds good, whether or not I merge the grad caching branch (I likely will as I feel the rest of the changes are sound and the grad...

@zerovl how does this compare to the grad caching #74 #65 impl that are being proposed? It'd be nice to do something a long the lines of the idea #74,...

@rom1504 there was some though for this in my refactoring, but I never made it fully realized, the is_master has a local arg set via args.log_local (or something similar), that...

@carlini I was actually wondering about this use case (multiple folders or datasets in different locations) and whether it'd work via command line or not... I was going to combine...

The latest changes have improved this for `open_clip` model with the modelling and loss code. It can be installed via `python setup.py install` or `pip -e .` from the git...

@iejMac @rom1504 what is the status? is the generalizable text transformer work in a good enough state to merge? The grad caching turned out to be quite a bit more...

> CI failing, some import error looks like PretrainedCfg needs to be defined as a dummy class for the typing arg in case where transformers isn't installed

FYI every weight at https://github.com/mlfoundations/open_clip/releases/tag/v0.2-weights has the first 8 char of the sha256 checksum after the last `-` ... which should be plenty for verifying corrupt downloads...