can't access python modules

[svedwards]

Hello Kern Lab -

I've just installed ReLERNN but am having trouble running the example file. It seems as if it can't find the appropriate python modules. The error looks like this:

Traceback (most recent call last):
File "/n/holylfs04/LABS/edwards_lab/Lab/sedwards/ReLERNN/ReLERNN/ReLERNN_SIMULATE", line 7, in
from ReLERNN.imports import *
File "/n/home06/sedwards/.local/lib/python3.10/site-packages/ReLERNN/init.py", line 3, in
from ReLERNN.imports import *
File "/n/home06/sedwards/.local/lib/python3.10/site-packages/ReLERNN/imports.py", line 4, in
import msprime as msp
File "/n/home06/sedwards/.local/lib/python3.10/site-packages/msprime/init.py", line 37, in
from msprime.ancestry import (
File "/n/home06/sedwards/.local/lib/python3.10/site-packages/msprime/ancestry.py", line 41, in
from . import demography as demog
File "/n/home06/sedwards/.local/lib/python3.10/site-packages/msprime/demography.py", line 43, in <modul$
import demes
File "/n/home06/sedwards/.local/lib/python3.10/site-packages/demes/init.py", line 9, in
from .demes import (
File "/n/home06/sedwards/.local/lib/python3.10/site-packages/demes/demes.py", line 13, in
from .load_dump import dumps as demes_dumps
File "/n/home06/sedwards/.local/lib/python3.10/site-packages/demes/load_dump.py", line 11, in
import ruamel.yaml
ModuleNotFoundError: No module named 'ruamel'
Traceback (most recent call last):

also farther down:

ModuleNotFoundError: No module named 'h5py'

Sorry if I am not providing enough information. I get a similar error when I run the singularity. Happy to provide additional info to help clarify the problem.

[svedwards]

I installed using the method listed on the web site. I made a conda environment then once in that environment I ran:

$ git clone https://github.com/kr-colab/ReLERNN.git
$ cd ReLERNN
$ pip install .

[andrewkern]

Hi @svedwards -- this seems like an installation issue. the h5py package should come along with tensorflow and the ReLERNN modules should just install with pip as you've tried.

I just tried out our suggested workflow on an ubuntu machine with a global install of tensorflow / cuda. The installation went like this

$ conda create -n relernn_test python=3.10 --yes
$ conda activate relernn_test
$ cd ReLERNN
$ which pip # confirm correct path to pip in my new conda env
$ pip install .
$ cd examples
$ ./example_pipeline.sh  # worked!

how was tensorflow installed on your machine? happy to help debug that too

cheers,
Andy

[svedwards]

Hi Andrew - thanks for the quick response. It very likely is my tensorflow installation - I think I was trying to use someone else's installation but obviously wasn't connecting with it properly. My protocol for installing tensorflow is here: https://docs.rc.fas.harvard.edu/kb/tensorflow/ but after the second step ('mamba create ...') it's complaining:
Looking for: ['python=3.10', 'pip', 'wheel']

[json.exception.parse_error.101] parse error at line 1, column 1: syntax error while parsing value - unexpected end of input; expected '[', '{', or a literal

I can check with my local computing staff and will also try the singularity container. I will check in tomorrow evening if I am still in trouble. Thank you!

[svedwards]

I'm working on a linux-64 machine

[andrewkern]

tensorflow / cuda installs can be a pain. let me know if you need recommendations on how to proceed.