fast-stable-diffusion
fast-stable-diffusion copied to clipboard
TheLastBen
After clicking on generate button on collab it goes to 100 but after completing it doesn't display the generated image on browser and generate button doesn't work tried everything restarting session tried on gradio server but nothing works
sometimes the servers are slow and cause the gradio interface to freeze, but that's only temporary
sometimes the servers are slow and cause the gradio interface to freeze, but that's only temporary
But without gradio server it happens again after you added dependencies it happened earlier I used it was completely alright??
And also image generated is visible in drive in output folder of stable diffusion but only it doesn't display in browser??
you probably maxed out your ram. Click on view resources and will probably see it in red. it seems to disconnect from the server if that happens. you have to reload the UI tab to reset it. try generating less in each batch or lower the resolution.
you probably maxed out your ram. Click on view resources and will probably see it in red. it seems to disconnect from the server if that happens. you have to reload the UI tab to reset it. try generating less in each batch or lower the resolution.
No iam connected to GPU its perfectly alright and iam running Collab on default settings
i mean the UI browser tab will disconnect from the colab browser tab if that makes sense. Hitting the generate button no longer works. Reload the tab and it will
i mean the UI browser tab will disconnect from the colab browser tab if that makes sense. Hitting the generate button no longer works. Reload the tab and it will bro i just waited for few minutes now it generated image but its very slow i don't know why earlier server was not that slow?
i mean the UI browser tab will disconnect from the colab browser tab if that makes sense. Hitting the generate button no longer works. Reload the tab and it will
is there any tip to make server fast?
sometimes the servers are slow and cause the gradio interface to freeze, but that's only temporary
how to make it fast
same happens to me, was happening also before but limited to "Extra" upscaling. Now it generates the first request and then in order to make it work again I need to, not only restart the webui, but also stop and restart the last step on colab to generate a new gradio server
same happens to me, was happening also before but limited to "Extra" upscaling. Now it generates the first request and then in order to make it work again I need to, not only restart the webui, but also stop and restart the last step on colab to generate a new gradio server
what could be the reason @TheLastBen
It consistently happens to me when I run a batch job and take the focus off tab the UI is open in, even if only for a few seconds. Once the batch finishes, it will not show in the UI and clicking the "Generate" button does nothing
It consistently happens to me when I run a batch job and take the focus off tab the UI is open in, even if only for a few seconds. Once the batch finishes, it will not show in the UI and clicking the "Generate" button does nothing
This is exactly how the issue presents for me as well. I think there's more to it than a slow server @TheLastBen. It doesn't make sense tab focus would impact that, does it?
fixed https://github.com/TheLastBen/fast-stable-diffusion/commit/cba1758085d03ad40c021e40582f4053f9cafe4a, use the colab from the main link : https://colab.research.google.com/github/TheLastBen/fast-stable-diffusion/blob/main/fast_stable_diffusion_AUTOMATIC1111.ipynb
@TheLastBen
fixed cba1758, use the colab from the main link : https://colab.research.google.com/github/TheLastBen/fast-stable-diffusion/blob/main/fast_stable_diffusion_AUTOMATIC1111.ipynb
I did a fresh run and got this:
2022-12-02 21:05:27.165115: I tensorflow/core/util/util.cc:169] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable TF_ENABLE_ONEDNN_OPTS=0.
LatentDiffusion: Running in eps-prediction mode
DiffusionWrapper has 859.52 M params.
Downloading100% 961k/961k [00:01<00:00, 720kB/s]
Downloading100% 525k/525k [00:01<00:00, 467kB/s]
Downloading100% 389/389 [00:00<00:00, 356kB/s]
Downloading100% 905/905 [00:00<00:00, 805kB/s]
Downloading100% 4.52k/4.52k [00:00<00:00, 3.79MB/s]
Downloading100% 1.71G/1.71G [00:21<00:00, 79.8MB/s]
Loading weights [90fd0a81] from /content/gdrive/MyDrive/AI/Awesome-Mix-v1.0.ckpt
Applying xformers cross attention optimization.
Model loaded.
Traceback (most recent call last):
File "/content/gdrive/MyDrive/sd/stable-diffusion-webui/webui.py", line 192, in
Failed to start.
fixed, I recommend not using A100 for inference as it is a waste of compute units, unless it's for production.
I have a similar problem. The generation can last only 10-20 seconds, but after that I have to wait at least as long, and at most several minutes, until one picture in the web interface is changed to a new one. I used to generate groups of images in high resolution and everything was displayed almost instantly when the generation was finished.
Checked. If "Batch count" = 1, then everything works as it should. It updates instantly, even on normal resolution pictures, and with a slight delay on high resolution pictures. But if "Batch count" > 1, the delay either increases nonlinearly or becomes infinite at high resolution. 512x512 and Batch count 2 pictures are updated. At 1024x1024+Highres fix and Batch count 2 I have already waited 10+ minutes and nothing has updated. Well, I will generate one picture at a time. UPD: I restarted and this time everything works perfectly, even with batches... With the repeatability of this problem will be difficult...