Closed qin-yu closed 4 months ago
Somehow cudatoolkit
11.1.74 from nvidia
channel is installed.
cudatoolkit 11.1.74 h6bb024c_0 nvidia
With the same conda and mamba installation, specifying cuda-version==12.3
would avoid the installation for package cudatoolkit
$ mamba env create -f ./environment-dev.yaml
name: plantseg-dev-cuda-12.3
channels:
- pytorch
- nvidia
- conda-forge
dependencies:
- python
- pydantic>2
- h5py
- zarr
- requests
- pyyaml
- scikit-image
- tifffile
- vigra
- cudnn
- pytorch
- pytorch-cuda=12.1
- cuda-version==12.3
- python-elf
- pyqt
- napari
- python-graphviz
cuda-cudart 12.1.105 0 nvidia
cuda-cupti 12.1.105 0 nvidia
cuda-libraries 12.1.0 0 nvidia
cuda-nvrtc 12.1.105 0 nvidia
cuda-nvtx 12.1.105 0 nvidia
cuda-opencl 12.4.127 0 nvidia
cuda-runtime 12.1.0 0 nvidia
cuda-version 12.3 h32bc705_3 conda-forge
cudnn 8.9.7.29 h092f7fd_3 conda-forge
conda
took more than 30 minutes to solve the environment, not done yet.
I'll update the dev env config.
It took 19 hours for conda
to solve but it's not done yet. Let's stick with mamba
Installation with the following method results in
cudatookit==11.1
, which I guess is incompatible withpytorch-cuda==12.1
and causes error during prediction.Let's try both
conda
andcuda-version==12.3