Is NVIDIA QUADRO M1000M supported?

[DevilaN]

Recently I've been trying to install SD from this repo on my friend's laptop.
Unfortunately his card has 2GB VRAM, so technically it should be possible to run SD with --lowvram flag, but should be doable.

Everything builds properly but python complains about not seeing graphic's card.

$ docker compose --profile auto up
[+] Running 1/0
 ⠿ Container webui-docker-auto-1  Created                                                                                                                                    0.0s
Attaching to webui-docker-auto-1
webui-docker-auto-1  | Mounted .cache
webui-docker-auto-1  | Mounted LDSR
webui-docker-auto-1  | Mounted BLIP
webui-docker-auto-1  | Mounted Hypernetworks
webui-docker-auto-1  | Mounted VAE
webui-docker-auto-1  | Mounted GFPGAN
webui-docker-auto-1  | Mounted RealESRGAN
webui-docker-auto-1  | Mounted Deepdanbooru
webui-docker-auto-1  | Mounted ScuNET
webui-docker-auto-1  | Mounted .cache
webui-docker-auto-1  | Mounted StableDiffusion
webui-docker-auto-1  | Mounted embeddings
webui-docker-auto-1  | Mounted ESRGAN
webui-docker-auto-1  | Mounted config.json
webui-docker-auto-1  | Mounted SwinIR
webui-docker-auto-1  | Mounted MiDaS
webui-docker-auto-1  | Mounted ui-config.json
webui-docker-auto-1  | Mounted BSRGAN
webui-docker-auto-1  | Mounted Codeformer
webui-docker-auto-1  | Mounted extensions
webui-docker-auto-1  | ++ nproc
webui-docker-auto-1  | + accelerate launch --num_cpu_threads_per_process=8 webui.py --listen --port 7860 --allow-code --lowvram --theme dark --enable-insecure-extension-access
webui-docker-auto-1  | The following values were not passed to `accelerate launch` and had defaults used instead:
webui-docker-auto-1  | 	`--num_processes` was set to a value of `1`
webui-docker-auto-1  | 	`--num_machines` was set to a value of `1`
webui-docker-auto-1  | 	`--mixed_precision` was set to a value of `'no'`
webui-docker-auto-1  | To avoid this warning pass in values for each of the problematic parameters or run `accelerate config`.
webui-docker-auto-1  | /usr/local/lib/python3.10/site-packages/torch/cuda/__init__.py:88: UserWarning: CUDA initialization: CUDA unknown error - this may be due to an incorrectly set up environment, e.g. changing env variable CUDA_VISIBLE_DEVICES after program start. Setting the available devices to be zero. (Triggered internally at ../c10/cuda/CUDAFunctions.cpp:109.)
webui-docker-auto-1  |   return torch._C._cuda_getDeviceCount() > 0
webui-docker-auto-1  | No module 'xformers'. Proceeding without it.
webui-docker-auto-1  | Warning: caught exception 'CUDA unknown error - this may be due to an incorrectly set up environment, e.g. changing env variable CUDA_VISIBLE_DEVICES after program start. Setting the available devices to be zero.', memory monitor disabled
webui-docker-auto-1  | Removing empty folder: /stable-diffusion-webui/models/BSRGAN
webui-docker-auto-1  | Checkpoint v1-5-pruned-emaonly.ckpt [cc6cb27103] not found; loading fallback sd-v1-5-inpainting.ckpt [c6bbc15e32]
webui-docker-auto-1  | LatentInpaintDiffusion: Running in eps-prediction mode
webui-docker-auto-1  | DiffusionWrapper has 859.54 M params.
webui-docker-auto-1  | Loading weights [c6bbc15e32] from /stable-diffusion-webui/models/Stable-diffusion/sd-v1-5-inpainting.ckpt
webui-docker-auto-1  | Applying cross attention optimization (InvokeAI).
webui-docker-auto-1  | Textual inversion embeddings loaded(0): 
webui-docker-auto-1  | Model loaded in 3.7s (0.6s create model, 3.0s load weights).
webui-docker-auto-1  | Running on local URL:  http://0.0.0.0:7860
webui-docker-auto-1  | 
webui-docker-auto-1  | To create a public link, set `share=True` in `launch()`.

I've tried whether nvidia-smi gives anything usefull and it seems that CUDA is there and works fine other than not working with SD.

root@b7727ec21cf4:/stable-diffusion-webui# nvidia-smi 
Wed Jan 25 12:00:03 2023       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 525.85.05    Driver Version: 525.85.05    CUDA Version: 12.0     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  Quadro M1000M       Off  | 00000000:01:00.0 Off |                  N/A |
| N/A   48C    P8    N/A /  N/A |      4MiB /  2048MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
+-----------------------------------------------------------------------------+

Currently I suspect that such device as m1000m is simply not supported by torch library, but maybe somebody would advice further investigation steps, please?

[AbdBarho]

Nvidia website says that the gpu has a compute capability of 5.0, which means it should in theory work with pytorch.
https://developer.nvidia.com/cuda-gpus

have you tried running pytorch without stable diffusion?

[DevilaN]

Sorry. I have no idea whatsoever about python related things as well pytorch etc. :(
Is there any command to test it or should I run some example files from some site with python inside container?

[DevilaN]

Something weird. It randomly is working and not. System related probably.

I've tested it with this script:

import torch

# setting device on GPU if available, else CPU
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
print('Using device:', device)
print()

#Additional Info when using cuda
if device.type == 'cuda':
    print(torch.cuda.get_device_name(0))
    print('Memory Usage:')
    print('Allocated:', round(torch.cuda.memory_allocated(0)/1024**3,1), 'GB')
    print('Cached:   ', round(torch.cuda.memory_reserved(0)/1024**3,1), 'GB')

Got following response:

root@b7727ec21cf4:/data# python test.py
Using device: cuda

Quadro M1000M
Memory Usage:
Allocated: 0.0 GB
Cached:    0.0 GB

Also this time SD is working properly.