Conda installation order / known good "recipe"?

[nbidler]

I have tried a number of times to install deepmd-kit on a Conda virtual environment. I also need to install cudatoolkit and cudnn.

However, each time I have tried this (even with multiple different version numbers) I keep getting "unsatisfiable conflict" errors.

Yes, even when following the conda create -n deepmd deepmd-kit=*=*gpu libdeepmd=*=*gpu lammps cudatoolkit=11.6 horovod -c https://conda.deepmodeling.com -c defaults guide, I have had these errors.

Does anyone have either

1. A list of packages in a working environment (must use CUDA)
2. A (shorter) list of what packages they installed, and in what order

Either would be appreciated.

[njzjz]

Please see #1362 and #1883.

[nbidler]

follow-up: tried using solution
conda create -n deepmd_tst deepmd-kit=*=*gpu libdeepmd=*=*gpu lammps-dp cudatoolkit=10.1 horovod -c https://conda.deepmodeling.org --dry-run
Success, installed without conflicts!

But oh no, a new error.

Traceback (most recent call last):
File "/share/apps/ana/envs/deepmd_tst/lib/python3.9/site-packages/numpy/init.py", line 328, in getattr
raise AttributeError("module {!r} has no attribute "
AttributeError: module 'numpy' has no attribute 'typeDict'
/share/apps/ana/envs/deepmd_tst/lib/python3.9/site-packages/tensorflow/python/framework/dtypes.py:511: FutureWarning: In the future np.object will be defined as the corresponding NumPy scalar.
np.object,
Traceback (most recent call last):
BIG LIST OF TRACE HERE
AttributeError: module 'numpy' has no attribute 'object'.
np.object was a deprecated alias for the builtin object. To avoid this error in existing code, use object by itself. Doing this will not modify any behavior and is safe.
The aliases was originally deprecated in NumPy 1.20; for more details and guidance see the original release note at:
https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations

Edit: there is more than one of this same error, I only put one here to save space.

I am going to attempt downgrading NumPy to 1.19 ( just with the command conda install numpy=1.19 to see if it accepts that first), will update with results

[nbidler]

The answer is "It creates a lot of conflicts" (see attached image) so I am going to try other solutions.
Might try conda create -n deepmd_tst2 deepmd-kit=*=*gpu libdeepmd=*=*gpu lammps-dp cudatoolkit=* horovod numpy=1.19 -c https://conda.deepmodeling.org to see if the additional restriction works

[nbidler]

current conflicts are:

  - feature:/linux-64::__glibc==2.17=0
  - cudatoolkit -> __glibc[version='>=2.17,<3.0.a0']
  - cudatoolkit -> libgcc-ng[version='>=11.2.0'] -> __glibc[version='>=2.17']
  - deepmd-kit[build=*gpu] -> __glibc[version='>=2.17']
  - deepmd-kit[build=*gpu] -> cudatoolkit[version='>=11.6,<12'] -> __glibc[version='>=2.17,<3.0.a0']
  - horovod -> __cuda
  - horovod -> __glibc[version='>=2.17']
  - horovod -> cudatoolkit[version='>=11.6,<12'] -> __glibc[version='>=2.17,<3.0.a0']
  - horovod -> nccl[version='>=2.14.3.1,<3.0a0'] -> __cuda[version='10.2|10.2.*|11.0|11.0.*|>=11.2,<12|11.1|11.1.*']
  - libdeepmd[build=*gpu] -> __glibc[version='>=2.17']
  - libdeepmd[build=*gpu] -> cudatoolkit[version='>=11.6,<12'] -> __glibc[version='>=2.17,<3.0.a0']
  - numpy=1.19 -> libgcc-ng[version='>=10.3.0'] -> __glibc[version='>=2.17']

Your installed version is: 2.17

I will try to determine which earlier versions lack this conflict.