-
Notifications
You must be signed in to change notification settings - Fork 374
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
scvi installing errors on get_num_classes and configure formatter 'console' #1540
Comments
Can you check what versions you have when you do?
|
Ah actually I see what's happening, it's downloading scvi-tools 0.14.5. This is the last release we had on bioconda https://anaconda.org/bioconda/scvi-tools Can you ensure that conda is not searching bioconda? You really want this (https://anaconda.org/conda-forge/scvi-tools), not the bioconda version. |
Hi, I actually have the same issue while installing scvi-tools with Error output after downloading and extracting packages:
Any advice? |
I can look into this, I ran across it as well but even with that message, it seems to have installed successfully |
Thanks for your quick response.
For torchmetrics
|
Yes, I am sure that my 2-time installations were made by using |
Actually more times with python3.8 and python3.9. |
I see, but as you can see from the output you pasted:
as far as I know there's a channel priority with conda, and you must have bioconda ahead of conda-forge, which is causing the issue. You can also install using pip just fine. |
Created a new environment again, manually changed channel priority to have conda-forge at the top before installing scvi-tools, still the same issue, unfortunately.
and
It does install in the end, however I get a lot of issues with packages later on, including when trying to later on install and run Spyder (Intel MKL FATAL ERROR: Cannot load libmkl_intel_thread.2.dylib.). With scvi and scanpy, I usually get the following: dlopen error with undefined symbol: H5Pset_fapl_ros3). I do not have enough knowledge to get around those problems (fairly new to python), unfortunately, but if you have any idea about what is going on then please let me know. Thank you for your time! |
I run MacOS Moneterey 12.3.1 with Intel graphics, if it helps |
Mine are 1.5.10 and 0.8.2, respectively |
If I run the following in this new environment
It shows:
Packages in env:
|
@AlinaKurjan can you try running Otherwise, I'd try a new environment, but pip install scvi-tools |
Seems to work now (with installing that anaconda h5py)! Will report if any issues spotted later on, thanks a lot for your help |
Hi thanks for the tips on the installation. I set the highest channel_priority to conda-forge shown as following.
Then, I installed successfully scVI with its version from conda-forge
But there was still sth wrong with torchmetrics when I imported scVI, shown as follows.
I searched online for solutions from NVIDIA/DeepLearningExamples#1113 by reinstalling a torchmetrics version=0.6.0. The installation problem is still unsolved. |
@2003100127 can you try with the linux-based operating system? There are some subtleties with installing on windows. |
Thanks a lot for all of your useful tips. The issue is fixed. It finally succeeded on Linux or Mac but it seems problematic with Windows. The Python version made for building a successful installation of scvi-tools is 3.8.13 in my case via the conda-forge channel. |
Hi,
I am one of the scvi users. When I installed the scvi tool in a conda env, I ended up with the below errors. I tested the installation in a conda env on Windows-based and Linux-based operating systems.
And furthurmore I got this below when I imported scvi in the jupyter notebook.
Please if there is any help with this issue. Thanks!
Best
The text was updated successfully, but these errors were encountered: