https://github.com/conda-forge/miniforge#miniforge3
This should limit duplication of downloads, hopefully
Environment Variable | Value |
---|---|
CONDA_ENVS_PATH |
D:\conda\conda-envs |
CONDA_PKGS_DIRS |
D:\conda\conda-packages |
Change
D:
to whatever drive you have on the workstation
conda
foldericacls D:\conda /grant:r Users:(OI)(CI)F /T
You can install brainreg and, for example, download the Allen Mouse Brain atlas (barrel-enhanced) at 10um resolution, with the following commands:
mamba create -n brainreg python==3.11 -y
conda activate brainreg
mamba install -y -c conda-forge brainreg
mamba install -y -c conda-forge pytables
Further documentation about brainreg, atlases and registration parameters can be found in the BrainGlobe's dedicated documentation.
Any atlas unavailable locally will be downloaded upon first call, so there is no need to pre-download atlases.
But this can be done like so, once brainreg is installed.
brainglobe install -a allen_mouse_bluebrain_barrels_10um
To avoid all users having the Atlases in their own user profile set the following environment variable
Environment Variable | Suggested value |
---|---|
BRAINGLOBE_CONFIG_DIR |
D:\conda\extras\brainglobe-atlases |
But it's not enough. After a first run of brainglobe (to show atlases for instance), a configuration file named bg_config.conf
will be created. Edit it in order to set the proper paths for where you want to download the atlases.
You can directly create upfront a text file named bg_config.conf
in this folder containing the paths you want to use:
[default_dirs]
brainglobe_dir = D:\conda\extras\brainglobe-atlases
interm_download_dir = D:\conda\extras\brainglobe-atlases
In order to avoid models being downloaded for each user, set CELLPOSE_LOCAL_MODELS_PATH
Environment Variable | Value |
---|---|
CELLPOSE_LOCAL_MODELS_PATH |
D:\conda\extras\cellpose-models |
mamba create -n cellpose-3111 python=3.8 -y
mamba activate cellpose-3111
mamba install pytorch pytorch-cuda=11.8 -c pytorch -c nvidia -y
pip install cellpose==3.1.1.1 scikit-image
Cellpose-SAM is not compatible with previous model (cyto3...).
mamba create -n cellpose-405 python=3.8 -y
mamba activate cellpose-405
mamba install pytorch pytorch-cuda=11.8 -c pytorch -c nvidia -y
pip install cellpose==4.0.5 scikit-image
As per this referenced issue, cellpose currently has a bug when it comes to training using --all_channels
. It cannot be used via the CLI.
This PR corrects the issue but has not been implemented. For now we can use the following to install a temporary version of Cellpose from our own GitHub fork
https://github.com/BIOP/cellpose/commit/3939d6a889debd1a6c73e5e2e2088a0ab7bb26eb
mamba create -n cellpose-308-fix python=3.8 -y
mamba activate cellpose-308-fix
mamba install pytorch pytorch-cuda=11.8 -c pytorch -c nvidia -y
pip install git+https://github.com/BIOP/cellpose.git@3939d6a889debd1a6c73e5e2e2088a0ab7bb26eb scikit-image
Follow the instructions in the ReadMe
Follow the instructions in the ReadMe
.yaml
file from the BIOP-Desktopconda env create -f path/to/env_omnipose.yaml
mamba create -n omnipose 'python==3.9.18'
mamba activate omnipose
mamba install pytorch torchvision pytorch-cuda=11.8 -c pytorch -c nvidia -y
pip install git+https://github.com/kevinjohncutler/omnipose.git@260e4d3f87a356d0d496e5db8febf80ccb2e48a7 aicsimageio torch_optimizer
Installing Pytorch via mamba seems to cause an issue, as we are calling Python and not mamba activate
from QuPath. The following installation works, but is messier.
mamba create -n omnipose python=3.9 -y
mamba activate omnipose
pip install git+https://github.com/kevinjohncutler/omnipose.git@260e4d3f87a356d0d496e5db8febf80ccb2e48a7 aicsimageio torch_optimizer scikit-image
pip uninstall torch torchvision
pip install torch torchvision --index-url https://download.pytorch.org/whl/cu118
mamba create -n empanada-023-gpu python=3.9 -y
mamba activate empanada-023-gpu
mamba install pytorch torchvision pytorch-cuda=11.8 -c pytorch -c nvidia -y
pip install "napari[all]" empanada-napari==0.2.3
A test image can be found ..\ptbiop-raw\public\0-Software\empanada\
conda create -n omero -c ome python=3.8 zeroc-ice36-python ipykernel
conda activate omero
python -m ipykernel install --user
conda install -c conda-forge omero-py
conda create -n ome_ngff python=3.9
conda activate ome_ngff
python.exe -m pip install --upgrade pip
pip install "napari[all]" napari-ome-zarr ome-zarr notebook
This ensures that the model weights are downloaded only once
Environment Variable | Value |
---|---|
SAMAPI_ROOT_DIR |
D:\conda\extras\sam-models |
mamba create -n samapi python=3.10 -y
mamba activate samapi
python -m pip install git+https://github.com/ksugar/samapi.git
pip uninstall torch torchvision
mamba install pytorch torchvision pytorch-cuda=11.8 -c pytorch -c nvidia
Activate SAM.bat
in the Public Desktop with the following codecall C:\ProgramData\miniforge3\Library\bin\mamba.bat activate samapi
uvicorn samapi.main:app --workers 2
pause
To be able to "pin" the shortcut to the task bar , right-click > properties > target , modifiy so cmd.exe /C "path-to-your-batch"
mamba create -n stardist-085-gpu python=3.7 cudatoolkit=10.1.* cudnn=7.6.* -c nvidia -y
mamba activate stardist-085-gpu
pip install tensorflow==2.3.* stardist==0.8.5
dataset to test
TODO
From instructions on github, for Windows make sure to install Build Tools for Visual Studio
mamba create -n spotiflow python=3.9.* pytorch torchvision pytorch-cuda=11.8 -c pytorch -c nvidia -y
mamba activate spotiflow
mamba install napari[all] pyqt
pip install spotiflow napari-spotiflow
mamba create --name tarrow python=3.8 -y
mamba activate tarrow
mamba install pytorch torchvision pytorch-cuda=11.8 -c pytorch -c nvidia -y
pip install git+https://github.com/weigertlab/tarrow.git
mamba create -n wsinfer pytorch torchvision pytorch-cuda=11.8 -c pytorch -c nvidia -y
mamba activate wsinfer
"mamba.exe" - Entry Point Not Found
To solve, please use the following commands :
conda config --set solver classic
conda update -n base -c conda-forge conda
conda config --set solver libmamba
conda install -f libarchive