Update Aurora/Sunspot environment

[gpauloski]

Here's a working Conda environment file for Aurora/Sunspot. By working I mean that the GPUs are accessible from within PyTorch on a Sunspot compute node.

Annoyingly, solving the environment, even with Mamba, had to be run overnight it was so slow :(.

I don't know how this will interact with LAMMPS so feel free to merge this or just close the PR and use it as reference.

I've included instructions in the file, but I leave them here as well.

1. On a login node, load modules with Conda.

   module use /soft/modulefiles/
   module load frameworks/2023.12.15.001

2. Build the Conda environment. Using Mamba may be wise.

   conda deactivate
   conda env create --file envs/environment-aurora.yml --force

3. Test the environment on a compute node.

   qsub -l select=1 -l walltime=60:00 -A <ALLOC> -q workq -I

   Here's a test Python file.

   import torch
   import intel_extension_for_pytorch as ipex
   
   print(f'Torch version: {torch.__version__}')
   print(f'GPU availability: {torch.xpu.is_available()}')
   print(f'Number of tiles = {torch.xpu.device_count()}')
   current_tile = torch.xpu.current_device()
   print(f'Current tile = {current_tile}')
   print(f'Curent device ID = {torch.xpu.device(current_tile)}')
   print(f'Device name = {torch.xpu.get_device_name(current_tile)}')

   Activate the environment.

   module use /soft/modulefiles/
   module load frameworks/2023.12.15.001
   conda activate mofa

   Run the test file.

   $ python test.py
   Torch version: 2.1.0a0+cxx11.abi
   GPU availability: True
   Number of tiles = 12
   Current tile = 0
   Curent device ID = <intel_extension_for_pytorch.xpu.device object at 0x147cd76f82e0>
   Device name = Intel(R) Data Center GPU Max 1550

[coveralls]

Pull Request Test Coverage Report for Build 8391791037

Details

• 0 of 0 changed or added relevant lines in 0 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage remained the same at 41.215%

---

Totals
Change from base Build 8380882788:	0.0%
Covered Lines:	4063
Relevant Lines:	9858

---

💛 - Coveralls

[WardLT]

This definitely gets us closer. I can recreate that Intel's test scripts work, run my model on CPU, but it fails on XPU. Thanks!