Since I fucked up and made this a new chain, we might as well use it as a "working commits" log. I'll start. Updated a few weeks ago, and 29fb5327640465fc83111e2170c5d8aa2b15266c seems to be working fine.

I just pulled today, it's fine. had to update lora extension then restart webui for it to work. I'm on >commit 93fad28a979727f9b1331dbdc447598824057cdc >Date: Tue Jan 24 21:13:05 2023 +0300

As I mentioned at >>1109 earlier, I recently jumped from >08c6f009a5ee92dd3218a942c08e8337c26352be to >7a14c8ab45da8a681792a6331d48a88dd684a0a9 because newer LoRAs weren't working with the version I was previously using. After git pulling, I had to follow the instructions below to get my installation working again: https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/7120#issuecomment-1404050224 I've only tested txt2img, prompt S/R, and native LoRA support so far, but it seems to be stable at the moment.

>>1204 I git pulled a few days ago to >2c1bb46c7ad5b4536f6587d327a03f0ff7811c5d and everything seems to be working fine without issue. I hear a few extensions don't work anymore, but I think those compatibility issues came up in earlier commits.

>>2691 Have you tried inpainting yet? I had issues with it on 1.0.0-pre, if that's fixed then it's time to upgrade

>>2787 Inpainting works fine. Here's a semi-low effort one I did of some official Taihou art.

commit 40e51fd6efa9c09a82c5ab391dbbd2c806971582 (HEAD -> master, origin/master, origin/HEAD) Date: Sat Feb 4 13:28:53 2023 +0300 add margin parameter to draw_grid_annotations >mostly works fine tested: txt2img, hires fix, img2img, image browser send to txt2img, lora extension lora extension doesn't read from generation metadata anymore. and model doesn't change either, regardless of setting? annoying. everything else seems fine.

>>3027 Thanks for reporting back. Gonna update

>>3027 >>3237 (me) "Send to" doesn't seem to work for me, and several options seem to have disappeared. Since "1.0.0-pre" is also broken I'm gonna revert to >>553

>>3284 That's really weird. I'm testing it right now, and "Send to" for txt2img to img2img works on my end. Which options disappeared for you? Did you do a git pull to the latest commit and then git checkout on the commit I posted? >>3151 Did you update the extension as well? I wonder what got changed that broke those fuctionalities.

>>3288 I had the same model switching issue as >>3151, to be precise. I'm on aa6e55e00140da6d73d3d360a5628c1b1316550d now after reading a related bug report (#7394), but it seems inpainting is back to producing artifacts outside of the mask on that commit.

>>3290 Ohh, I've had the model switching option set to "do not change" this whole time, so I didn't even notice that bug. That's annoying. Lora extension on my end still reads from metadata and adds the right models and weights when I send from PNG Info to txt2img, at least.

After having rebooted my PC since posting, even inpainting now works in aa6e55e00140da6d73d3d360a5628c1b1316550d.

posting a commit date with the commit hash helps to easily see how recent a commit is

>>3422 Glad to hear that you were able to find a good commit after all! >>3473 >>2691 Good idea, let me add the link to the commit that I'm on: https://github.com/AUTOMATIC1111/stable-diffusion-webui/commit/2c1bb46c7ad5b4536f6587d327a03f0ff7811c5d The commit was made on January 30th, 2023.

>>3501 Thanks for reporting that initial commit, it set me in the right direction! >> 3473 Fair enough. Here's my full info: >commit aa6e55e00140da6d73d3d360a5628c1b1316550d (HEAD) >Author: AUTOMATIC <16777216c@gmail.com> >Date: Sun Jan 29 11:53:05 2023 +0300 > do not display the message for TI unless the list of loaded embeddings changed

>>7723 I was able to get 0cc0ee1 working after a restart (thank you >>3422 for reminding me), and >>7720 corroborates that it is a functional commit. https://github.com/AUTOMATIC1111/stable-diffusion-webui/commit/0cc0ee1bcb4c24a8c9715f66cede06601bfc00c8 The commit was made on February 20th, 2023.

was 13 days out of date. just pulled and everything I've tested so far is working well. >git log -n 1 commit 0cc0ee1bcb4c24a8c9715f66cede06601bfc00c8

How are the latest versions doing? It's been a full month since my last pull, heard some bad news regarding recent Gradio changes breaking many things.

>>14795 The new gradio version did break a lot of stuff, but most of the major problems seem to have been dealt with. It's still kind of unstable on mobile, so if you access it from your phone a lot (which I do) you might want to avoid updating. I've managed to get mine to work on mobile by adding "queue=False" to the buttons that send PNG info to other tabs (otherwise the app hangs and you have to refresh the page), and by adding the below to user.css to disable the loading animation for settings dropdowns, which still seems to get stuck a lot on mobile: [id^="setting_"] > div[style*="position: absolute"] { display: none !important; }

>>15875 Same Anon, I've more recently realized I can fix most of the problems with mobile by adding the arg "--no-gradio-queue". The name is misleading, even with this on requests still get queued, it just forces the browser to use http polling instead of websockets to keep track of progress, which seems to prevent various problems where the UI stops working or stuff never finishes loading. I suspect it's a crappy implementation of the websockets stuff in js that has race conditions (like if the page loads slower than some recv callback fires an event gets missed), or that lacks some kind of logic to reconnect dead websockets, or something. With websockets disabled the mobile version works again. Really wish I'd known about this arg sooner.