Support for using cloud GPU

[DataTracer]

I want to use the SAM on Qupath but don't have any GPU on my computer. Would it be possible to provide access to something similar to Cloud collab

[ksugar]

Hi @DataTracer, thank you for the question.

Unfortunately, Google Colab does not allow to host a web service offerings not related to interactive computing with Colab.
https://research.google.com/colaboratory/faq.html#limitations-and-restrictions

I don't have any good suggestions for free GPU options. For paid options, we can set up the samapi server with a GPU instance on a public cloud service (e.g. Amazon AWS, Microsoft Azure, Google Cloud).

Did you try running the samapi server without GPU (CPU mode) on your machine? I have tested samapi without GPU, and it is still feasible to work with it.

The installation is simple as below:

conda create -n samapi -y python=3.10
conda activate samapi
python -m pip install git+https://github.com/ksugar/samapi.git

Then you can run it with the following command.

export PYTORCH_ENABLE_MPS_FALLBACK=1 # Required for running on Apple silicon
uvicorn samapi.main:app

I hope it works for you.

[DataTracer]

Thanks! Your extension works really well and I tried it on CPU but obviously not as fast :-)
I am trying to train the classifiers of Qupath on some users DAB data, fingers crossed!
Thanks for building this and making it available for the community

Best wishes
Nabanita

[DataTracer]

Hi, I am using a workstation with NVIDIARTX5000 and
installed as below :
If you're using a computer with CUDA-compatible GPU, install cudatoolkit.

conda install -c conda-forge -y cudatoolkit=11.7
If you're using a computer with CUDA-compatible GPU on Windows, install torch with GPU-support with the following command.

Windows with CUDA-compatible GPU only

python -m pip install "torch>=1.13.1,<2.0" torchvision --index-url https://download.pytorch.org/whl/cu117

Then when i am launching the server i get the following message

I could be wrong but RTX 5000 is CUDA compatible ?

Looking forward to your reply
Nabanita

[ksugar]

Hi @DataTracer, RTX 5000 is compatible with CUDA 11.7 according to this matrix.
https://forums.developer.nvidia.com/t/cuda-compatibility-between-nvidia-rtx-a5000-and-geforce-rtx-4060-ti/264216

Could you share the output of the nvidia-smi command? The Nvidia driver may need to be updated.

Example:

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 528.89       Driver Version: 528.89       CUDA Version: 12.0     |
|-------------------------------+----------------------+----------------------+
| GPU  Name            TCC/WDDM | Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  NVIDIA RTX A4000   WDDM  | 00000000:01:00.0 Off |                  Off |
| 41%   31C    P8     7W / 140W |    238MiB / 16376MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+

[DataTracer]

Hi @ksugar , Thank you, I am trying to run the nvidia-smi command, it disappears instantly, so I am unable to share the output.

On a separate note , I also posted this on image.sc but do you have any thoughts on how to use SAM annotations to train the object classifier in QuPath?

Thanks a lot
Best
Nabanita