[Bug]: Looks like some of the images don't upload correct

[civitai]

[Bug]: Looks like some of the images don't upload correct

[JustMaier]

We've put a check in place to let users know when their image failed to upload but we're still not completely sure why they're failing to upload occasionally.

Hopefully we've stemmed the bleeding.

[WASasquatch]

We've put a check in place to let users know when their image failed to upload but we're still not completely sure why they're failing to upload occasionally.

Hopefully we've stemmed the bleeding.

It does seem like they are all uploading at once? Which would mean each upload is relying on a CPU thread. If they upload say 12 images, with only a 4 core processor and 8 threads, and then a bunch of PC activity, I'd assume uploads could fail rather often for such users. Same for multi-threaded downloaders and people reporting failed downloads, to find out they're setting 8 threads on a CPU with only 8 threads, and the OS, browser, etc, to allocate to.

Edit: 2/21/2023:

Seems like something is going on now, where images just won't upload. Just getting Processing... for minutes and minutes and they never upload. No activity either on net for the process.

Edit #2

Seems that's just cause the whole site is down.

[WASasquatch]

Still having massive issues all the time guys.

Biggest issue it saying images are complete, when they are not, to then see your model is filled with broken images. Not it's failing to upload any images to uploaded model.

[WASasquatch]

Any word on this? It still happens every upload. Any more then one upload and they pretty much all fail. One at a time works fine. If you made the JS upload all queued sequentially it would probably fix this for everyone. Maybe have a button for multi-threaded.

Though it seems the uploader script has issues, I don't see this anywhere else, and my CPU (13500kf) and net is more than capable.

[JustMaier]

I assume this is failing due to the client-side image nsfw and face detection we run on all of the images. We plan on rolling out a new system that will replace this very soon, which should fix this.