Creating conda environment using yml file

@MARTIN_DURANT Hello! I have been strugling create new conda environment for my project for two weeks

Now my problem is that I need to install previous versions of some packages using latest packages I can build the environment but using past packages the environment creation from yesterday still spin on solving environment:
and until now no progress or error
I can not understand why that long

This is thw yml file

name: null

channels:
  - pytorch
  - conda-forge
  - defaults

dependencies:
  - bokeh
  - captum
  - ccache
  - cmake
  - cudnn=8.2
  - cudatoolkit=11.3
  - cudatoolkit-dev=11.3
  - cupy
  - gxx_linux-64=9.4
  - dask
  - dask-ml
  - dask-labextension
  - datashader
  - datasets
  - featuretools
  - gh
  - git
  - h5py
  - hvplot
  - imbalanced-learn
  - jupyterlab
  - jupyterlab-git
  - jupyterlab-lsp
  - jupyterlab-nvdashboard
  - lightgbm
  - matplotlib
  - mpi4py # installs cuda-aware openmpi
  - nccl
  - numba
  - openmpi
  - optuna
  - pandas
  - panel
  - einops
  - sacrebleu
  - prettytable
  - pip
  - pip:
    - -r requirements.txt
  - pyarrow
  - python=3.8
  - tokenizers=0.10.3
  - sentencepiece
  - python-language-server
  - pytorch=1.9.0
  - torchaudio=0.9.0
  - torchvision=0.10.0
  - tqdm=4.49.0
  - transformers=4.8.2
  - pyviz_comms
  - scikit-learn
  - scipy
  - shap
  - tensorboard
  - torchtext
  - umap-learn
  - wandb
  - xarray
  - xeus-python
  - xgboost

and this is the requirements.txt file:

pandas_bokeh
horovod[pytorch]
ray

# make sure horovod is re-compiled if environment is re-built
--no-binary=horovod        

Any explanation why the environment can not be build?

Sorry, I am not able to add more information beyond what was written in the linked issue.

The issue now with conda building it just stop at solving environment and no progress is that normal?

I suppose it’s not normal, but I am really not an expert on this