Replies: 15 comments 7 replies
-
I will only say i can use this DB repo's with a 1080 TI without problems: https://github.com/ShivamShrirao/diffusers/tree/main/examples/dreambooth (WSL + Ubuntu) And my config in my Train file looks like this:
|
Beta Was this translation helpful? Give feedback.
-
Also related to: |
Beta Was this translation helpful? Give feedback.
-
I am having this same problem. GTX 1080Ti Windows 10 Error no kernel image is available for execution on the device at line 167 in file D:\ai\tool\bitsandbytes\csrc\ops.cu |
Beta Was this translation helpful? Give feedback.
-
I have the same issue. |
Beta Was this translation helpful? Give feedback.
-
Is there any possibility that might be fixable ? I would love to avoid buying a RTX 3090 / 4090 to run this. Thank you |
Beta Was this translation helpful? Give feedback.
-
I'm using Titan Xp, and i have the same issue here |
Beta Was this translation helpful? Give feedback.
-
that path D:\ai\tool\bitsandbytes\csrc\ops.cu is hardcoded somewhere, we just need to figure out where (which file) and point it ot the local bitsandbytes |
Beta Was this translation helpful? Give feedback.
-
I did find a reference to D:\ai\tool\bitsandbytes\csrc\ops.cu in K:\Webui\stable-diffusion-webui\extensions\sd_dreambooth_extension\bitsandbytes_windows\libbitsandbytes_cuda116.dll. Could even change the drive letter, but, there's no ops.cu located anywhere in the webui installation. Besides, I think it's just a error message, not actually calling that file. But, I know nothing about these python things. |
Beta Was this translation helpful? Give feedback.
-
You finally fixed it, 1080 TI working now, how you did it? |
Beta Was this translation helpful? Give feedback.
-
I'd imagine it was compiled on a different setup, and dependencies and such just, were in the right places? Something like that seemed to be the issue. Not a coder. But can confirm, managed to train with Dreambooth for some time before my 1080ti died. Unrelated causes. 🙂 |
Beta Was this translation helpful? Give feedback.
-
solved this by seting use_8bit_adam = 0 |
Beta Was this translation helpful? Give feedback.
-
Also worked for me after disabling 8-bit-adam. I was trying to train Lora on the Kohya-ss gui, and for the past 3 days was trying to find some solution and I found it today, finally. Running on 1060 gtx 6gb |
Beta Was this translation helpful? Give feedback.
-
I'm using a 1080 graphic card, had the same issue. Made it running just now. Don't use the "AdamW8bit" optimizer (I randomly picked Lion, might work with other options haven't been tested yet). At least it's running: |
Beta Was this translation helpful? Give feedback.
-
Thanks everyone in this discussion. I have few problems include OP problem. Just in case put here my way to fix it. "Okay, I solved the problem. Pytorch downloaded by itself after a restart. Then there were problems in the "sample" folder (it's easy to delete). And at the end there was a problem due to the fact that I have a 1080ti. I'm writing just in case it helps someone." |
Beta Was this translation helpful? Give feedback.
-
This is a centralized place for folks with a 1080TI Family GPU to discuss issues related to that specific card and the above error below.
As this is hardware-specific, I'll try to redirect any related issues here.
My best guess is that this is related to 8bit adam or xformers, as both are using precompiled libraries which probably require specific code for this family of GPU - but that's just a guess.
#52
Beta Was this translation helpful? Give feedback.
All reactions