nvidia-smi.exe - The system cannot find the file specified causing timeout

[lodha13]

Thanks for sharing the code.

As per instruction I deployed the code in my local machine (32GB ram, 16GB GPU),

I started like a charm (tough it took some time to download all the stuff) but when I enterred some prompt it failed saying COULDN'T START at the step

Denoising image in 20 steps on cpu (unchunked)

what am i doing wrong ?

[painebenjamin]

Hello,

Glad the installation process was smooth, that's had a lot of work recently so it's good to know it's working. Sorry you're having problems with the backend - it's not immediately clear what's going on just from that error alone, though.

First let's just establish a baseline - make sure the error returns when you exit and start the server again, and give your browser a cache clear and refresh. If that doesn't make the issue go away, then it'd be very helpful to see your logs. Log files are located at ~/.cache/enfugue.log, where ~ is your home directory (\Users\<Your Name> on Windows, /home/<Your Name> on Linux). There could be multiple log files in that directory with names starting with enfugue, all of them would be useful.

You may also see some errors in the developer console of your browser - for most browsers that can be accessed by pressing F12. It would be useful to see those as well.

[lodha13]

enfugue.log
2024-01-09 00:33:34,689 [enfugue] ERROR (gpu.py:145) Couldn't execute nvidia-smi (binaryC:\Program Files\NVIDIA Corporation\NVSMI\nvidia-smi.exe`): [WinError 2] The system cannot find the file specified

2024-01-09 00:33:44,679 [enfugue] ERROR (gpu.py:145) Couldn't execute nvidia-smi (binary C:\Program Files\NVIDIA Corporation\NVSMI\nvidia-smi.exe): [WinError 2] The system cannot find the file specified

2024-01-09 00:34:36,946 [pibble] ERROR (base.py:403) Unexpected exception in application: OSError() Process died while waiting for result.`

enfugue_engine.log

2024-01-09 00:30:40,422 [urllib3.connectionpool] DEBUG (connectionpool.py:546) https://huggingface.co:443 "HEAD /openai/clip-vit-large-patch14/resolve/main/config.json HTTP/1.1" 200 0
2024-01-09 00:30:42,071 [enfugue] DEBUG (process.py:374) Instruction 6 beginning task “Executing Inference”
2024-01-09 00:30:42,071 [enfugue] DEBUG (manager.py:5019) Calling pipeline with arguments {'latent_callback': 'function', 'width': '512', 'height': '512', 'tile': '(False, False)', 'num_results_per_prompt': '1', 'tiling_vae': 'False', 'tiling_unet': 'False', 'tiling_stride': '0', 'prompts': '[Prompt]', 'progress_callback': 'function', 'latent_callback_steps': '5', 'latent_callback_type': 'pil'}
2024-01-09 00:30:42,071 [enfugue] DEBUG (pipeline.py:3644) Calculated overall steps to be 21. 0 image prompt embedding probe(s) + 20 UNet step(s) (1 spatial chunk(s) * 1 temporal chunk(s) * 20 inference step(s) * 1 denoising iteration(s)) + 1 VAE step(s) (1 chunk(s) * (0 encoding step(s) + (1 decoding step(s) * (1 frame(s) / 1 frame(s) per decode))))
2024-01-09 00:31:27,479 [enfugue] DEBUG (pipeline.py:2269) Denoising image in 20 steps on cpu (unchunked)

[painebenjamin]

Hey @lodha13, ENFUGUE isn't seeing your GPU. Are you on AMD?

[lodha13]

its intel i7, though i am on intel i7 but what i see is below (amd in dir name), not sure where it came from C:\Windows\System32\DriverStore\FileRepository\nvdimm.inf_amd64_57f50f6dd441f8cd nvdimm.inf_amd64_57f50f6dd441f8cd I dont have nvidia-smi.exe in my laptio, does it not work without that, I tried installing CUDA from nvidia but that also didnt install nvidia-smi.exe

[lodha13]

One thing I tried in different github codebase is below, and that successfully generated the image.
COMMANDLINE_ARGS=--skip-torch-cuda-test --no-half --lowvram --opt-split-attention --xformers
https://github.com/AUTOMATIC1111/stable-diffusion-webui

How can i do similar thing here, didnt find something similar.
can you please help, how to resolve this issue, looks like I dont have nvidia installed, can it not just use CPU instead of GPU. though I see in task manager GPU is shown

[painebenjamin]

Hello, sorry for the delay, I've been very busy with DragNUWA!

ENFUGUE will run on CPU-only, although it is very slow and many features will not work. I just tested and was able to generate an image without using my GPU, though there may be more things going on here.

How long does it take for you to receive the COULDN'T START notification? Is it immediate, or does it take ~1-3 minutes?

[lodha13]

when i started first it took quite some time, but after that it immediately says.

Is there any alternative where it doesnt need below file

2024-01-10 13:54:00,819 [enfugue] ERROR (gpu.py:145) Couldn't execute nvidia-smi (binary C:\Program Files\NVIDIA Corporation\NVSMI\nvidia-smi.exe): [WinError 2] The system cannot find the file specified

I have also tried a fresh install with conda (earlier i had selected portable option where it did all stuff) but with aslo also i keep on getting the error in log that the file is not present.

Since i dont have nvidia driver i think i dont have that file at that location. How can i generate an image without this file

[lodha13]

Sir, please help, i really wanted to try this amazing tool. any help is appreciated

[painebenjamin]

Hello @lodha13,

I am finishing builds of 0.3.3 which will include an option for you to increase the timeout duration. I believe that doing so will allow you to make images; I'll post again in this thread when it is released so you're aware.

[lodha13]

I think the problem is not timeout, the problem is its not finding nvidia-smi.exe and it keeps on checking fir this file until timeout. Sent from Yahoo Mail on Android On Fri, Jan 19, 2024 at 2:17 AM, Benjamin ***@***.***> wrote: Hello @lodha13, I am finishing builds of 0.3.3 which will include an option for you to increase the timeout duration. I believe that doing so will allow you to make images; I'll post again in this thread when it is released so you're aware. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: ***@***.***>