CUDA incompatibility when running docker compose up for SAM2 image

[aubenoire17]

I've been following the Quickstart guide here - https://github.com/HumanSignal/label-studio-ml-backend?tab=readme-ov-file#quickstart

git clone https://github.com/HumanSignal/label-studio-ml-backend.git
cd label-studio-ml-backend/label_studio_ml/examples/{MODEL_NAME}
docker-compose up

after running docker compose up, the image gets built but halfway it runs into this error:

=> ERROR [segment_anything_2_video stage-0 5/17] RUN conda install -c "nvidia/label/cuda-12.1.1" cuda -y 6.4s

[segment_anything_2_video stage-0 5/17] RUN conda install -c "nvidia/label/cuda-12.1.1" cuda -y:
1.115 /opt/conda/lib/python3.10/site-packages/conda/base/context.py:201: FutureWarning: Adding 'defaults' to channel list implicitly is deprecated and will be removed in 25.3.
1.115
1.115 To remove this warning, please choose a default channel explicitly with conda's regular configuration system, e.g. by adding 'defaults' to the list of channels:
1.115
1.115 conda config --add channels defaults
1.115
1.115 For more information see https://docs.conda.io/projects/conda/en/stable/user-guide/configuration/use-condarc.html
1.115
1.115 deprecated.topic(
2.074 Channels:
2.074 - nvidia/label/cuda-12.1.1
2.074 - defaults
2.074 Platform: linux-64
2.074 Collecting package metadata (repodata.json): ...working... done
4.053 Solving environment: ...working... failed
5.799
5.799 LibMambaUnsatisfiableError: Encountered problems while solving:
5.799 - nothing provides __win needed by cuda-12.4.1-h382c6e5_0
5.799
5.799 Could not solve for environment specs
5.799 The following packages are incompatible
5.799 ├─ cuda =* * is installable with the potential options
5.799 │ ├─ cuda 12.1.1 would require
5.799 │ │ └─ cuda-toolkit >=12.1.1 * with the potential options
5.799 │ │ ├─ cuda-toolkit 12.1.1 would require
5.799 │ │ │ └─ cuda-libraries >=12.1.1 * with the potential options
5.799 │ │ │ ├─ cuda-libraries 12.1.1 would require
5.799 │ │ │ │ └─ libcublas >=12.1.3.1 *, which can be installed;
5.799 │ │ │ └─ cuda-libraries 12.4.1 would require
5.799 │ │ │ └─ libcublas =12.4.5.8 , which can be installed;
5.799 │ │ ├─ cuda-toolkit 12.4.1 would require
5.799 │ │ │ └─ __win = *, which is missing on the system;
5.799 │ │ └─ cuda-toolkit 12.4.1 would require
5.799 │ │ └─ cuda-libraries =12.4.1 , which can be installed (as previously explained);
5.799 │ ├─ cuda 12.4.1 would require
5.799 │ │ └─ __win = *, which is missing on the system;
5.799 │ └─ cuda 12.4.1 would require
5.799 │ └─ cuda-toolkit =12.4.1 * with the potential options
5.799 │ ├─ cuda-toolkit 12.4.1, which cannot be installed (as previously explained);
5.799 │ └─ cuda-toolkit 12.4.1, which can be installed (as previously explained);
5.799 └─ pytorch-cuda =12.1 * is not installable because it requires
5.799 └─ libcublas >=12.1.0.26,<12.1.3.1 *, which conflicts with any installable versions previously reported.