Open ryao-mdanderson opened 7 months ago
Hi ryao-mdanderson,
Thank you for spotting a typo - https://github.com/ventolab/CellphoneDB/blob/master/README.md is now fixed. The tutorial you need is in: https://github.com/ventolab/CellphoneDB/blob/master/notebooks/T0_DownloadDB.ipynb
Best,
Robert.
Hello Robert @datasome
Thank you for your fix. I am not familiar with Jupyter notebook, in order to use CellphoneDB, should I have to use the installation instruction 4. Set up the kernel for the Jupyter notebooks? Should I have to download database using notebooks/T0_DownloadDB.ipynb?
I am an HPC system admin, I plan to set up a module for cluster users. The Jupyter notebooks approach seems to be a personal space, I wonder if this is suitable for cluster setting? Thank you for your suggestion.
Best, Rong
Hi Rong,
The answer to both questions is "yes". Personally, I don't have experience of running Jupyter notebooks in a HPC environment, but it's certainly possible and done frequently - here are a few example links I found:
In our institute we have something called JupyterHub, I understand it allows users to run Jupyter notebooks on a remote cluster/VM from the comfort of their browser. This to me seems to be an ideal set up for the user - that allows for easy tweaking of parameters and re-running analyses without having to worry about HPC job management layer.
@prete, is there any high level guidance you could provide on the use of Jupyter notebooks in an HPC environment?
Best,
Robert.
Hi Robert @datasome,
I appreciate your answers and providing me links. This is helpful.
We do have Jupyter notebook (multiple versions as module) set up on HPC; Our cluster users are able to launch Jupyter.
For CellphoneDB, I use a service account to install the conda environment 'cpdb-5.0.0'.
From instruction, 4. Set up the kernel for the Jupyter notebooks. 1)Install the ipython kernel: pip install -U ipykernel. (cpdb-5.0.0) [ris_hpc_apps@ldragon4 ~]$ pip install -U ipykernel 2) Add the environment as a jupyter kernel: python -m ipykernel install --user --name 'cpdb'. (cpdb-5.0.0) [ris_hpc_apps@ldragon4 ~]$ python -m ipykernel install --user --name 'cpdb'
I notice the command installed kernelspec cpdb in /risapps/ris_hpc_apps/.local/share/jupyter/kernels/cpdb which is in the service account .local folder.
3) Open/Start Jupyter and select the created kernel. I am puzzled in this step, how cluster user to do this step with the cpdb jupyter kernel installed in other accoount .local? Is the application desired for a single user mode? i.e. install conda environment and Jupyter kernel in user space?
Thank you! Rong
Hi Rong,
The default would be that the user installs the kernel for themselves, but I suspect there are ways of users using shared kernels - I just don't have experience of setting that up. Perhaps the following links might help:
Sorry I could not be of more help.
Best,
Robert.
Dear Robert @datasome ,
No worries, your help is already much appreciated!
Best, Rong
Dear CellphoneDB authors,
The README "Installing CellphoneDB " section step 5. download the database, Follow this tutorial, the link is 404 page not found. Please advise how to download database.
Thank you.