Package versions

[esb5324]

Hello, I have been having trouble setting up a conda environment with the required packages. Is there any flexibility with any of the versions, so I can try and see if that works? I have been getting errors where it seems like the versions are in conflict with each other. Thank you for your advice! Elle

[jaydu1]

Hi, can u provide the platform: win/linux/maxos? I am aware of some package conflicts on wins using conda-forge.

You can use inexact version requirements, e.g. conda install 'matplotlib>=1.4.3' to try to avoid the some of the conflicts.

[esb5324]

Hi Jinhong,

Thanks for your quick response. I am using Linux.

Elle

---

From: Du Jinhong @.> Sent: Wednesday, October 18, 2023 8:18 PM To: jaydu1/scVAEIT @.> Cc: Tang, Elle Salina @.>; Author @.> Subject: Re: [jaydu1/scVAEIT] Package versions (Issue #4)

You don't often get email from @.*** Learn why this is importanthttps://aka.ms/LearnAboutSenderIdentification

Hi, can u provide the platform: win/linux/maxos? I am aware of some package conflicts on wins using conda-forge.

You can use inexact version requirements, e.g. conda install 'matplotlib>=1.4.3' to try to avoid the some of the conflicts.

— Reply to this email directly, view it on GitHubhttps://github.com/jaydu1/scVAEIT/issues/4#issuecomment-1769673103, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ANQQ3YM4CBHJDI4I2HRERD3YABWT7AVCNFSM6AAAAAA6GH6LMOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTONRZGY3TGMJQGM. You are receiving this because you authored the thread.Message ID: @.***>

[jaydu1]

The main dependencies are tensorflow and tensorflow-probability.

I just tested on the updated versions of these packages:

# Name                    Version                   Build  Channel
h5py                      3.9.0           nompi_py311h3839ddf_103    conda-forge
numpy                     1.26.0          py311h64a7726_0    conda-forge
pandas                    2.1.1           py311h320fe9a_1    conda-forge
python                    3.11.5          hab00c5b_0_cpython    conda-forge
scikit-learn              1.3.1           py311hc009520_0    conda-forge
scipy                     1.11.3          py311h64a7726_0    conda-forge
tensorflow                2.12.1          cuda112py311h63ba870_0    conda-forge
tensorflow-base           2.12.1          cuda112py311hb0f8ee9_0    conda-forge
tensorflow-estimator      2.12.1          cuda112py311h9aaad64_0    conda-forge
tensorflow-gpu            2.12.1          cuda112py311h0bbbad9_0    conda-forge
tensorflow-probability    0.20.0             pyhd8ed1ab_0    conda-forge

It seems to also work. Let me know if you still encouter any problem.

[jaydu1]

I have formally tested it on the updated packages. You can follow the instructions here to create a conda environment (if you don't have mamba, simply replace it by conda):

mamba create --name tf python=3.9 -y
conda activate tf
mamba install -c conda-forge "tensorflow>=2.12" "tensorflow-probability>=0.12" pandas jupyter -y
mamba install -c conda-forge "scanpy>=1.9.2" matplotlib scikit-learn -y

The UMAP visualizations will be slightly effected by scanpy and umap-learn version, though they are similar to those in the paper.

[esb5324]

Thank you! I had to upgrade numpy to 1.22.4 + and run pip install tensorflow-addons, and then it worked.

Elle Tang Pronouns: she/her/hers Statistics Ph.D. student

---

From: Du Jinhong @.> Sent: Saturday, October 21, 2023 9:30 AM To: jaydu1/scVAEIT @.> Cc: Tang, Elle Salina @.>; Author @.> Subject: Re: [jaydu1/scVAEIT] Package versions (Issue #4)

You don't often get email from @.*** Learn why this is importanthttps://aka.ms/LearnAboutSenderIdentification

I have formally tested it on the updated packages. You can follow the instructions here to create a conda environment (if you don't have mamba, simply replace it by conda):

mamba create --name tf python=3.9 -y conda activate tf mamba install -c conda-forge "tensorflow>=2.12" "tensorflow-probability>=0.12" pandas jupyter -y mamba install -c conda-forge "scanpy>=1.9.2" matplotlib scikit-learn -y

The UMAP visualizations will be slightly effected by scanpy and umap-learn version, though they are similar to those in the paper.

— Reply to this email directly, view it on GitHubhttps://github.com/jaydu1/scVAEIT/issues/4#issuecomment-1773793050, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ANQQ3YN4QMFAIP4EEPGHZSLYAPE7TAVCNFSM6AAAAAA6GH6LMOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTONZTG44TGMBVGA. You are receiving this because you authored the thread.Message ID: @.***>