Error in Installation Modulus v22.03

[Spartacus]

Hi:
I am trying to install Nvidia Modulus v22.03 on a cluster with a Conda environment. After installation, the job helmholtz.py cannot run with the following error, I wish to ask how to solve this:

Traceback (most recent call last):
File “/project/user/DL/anaconda3/envs/modulus2203/lib/python3.9/site-packages/modulus-22.3-py3.9.egg/modulus/distributed/manager.py”, line 124, in initialize
DistributedManager.initialize_env()
File “/project/user/DL/anaconda3/envs/modulus2203/lib/python3.9/site-packages/modulus-22.3-py3.9.egg/modulus/distributed/manager.py”, line 67, in initialize_env
rank = int(os.environ.get(“RANK”))
TypeError: int() argument must be a string, a bytes-like object or a number, not ‘NoneType’

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File “/project/user/DL/modulus2203/examples/helmholtz/helmholtz.py”, line 95, in
run()
File “/project/user/DL/anaconda3/envs/modulus2203/lib/python3.9/site-packages/modulus-22.3-py3.9.egg/modulus/hydra/utils.py”, line 58, in func_decorated
DistributedManager.initialize()
File “/project/user/DL/anaconda3/envs/modulus2203/lib/python3.9/site-packages/modulus-22.3-py3.9.egg/modulus/distributed/manager.py”, line 127, in initialize
DistributedManager.initialize_slurm(port)
File “/project/user/DL/anaconda3/envs/modulus2203/lib/python3.9/site-packages/modulus-22.3-py3.9.egg/modulus/distributed/manager.py”, line 108, in initialize_slurm
DistributedManager.setup(
File “/project/user/DL/anaconda3/envs/modulus2203/lib/python3.9/site-packages/modulus-22.3-py3.9.egg/modulus/distributed/manager.py”, line 150, in setup
os.environ[“MASTER_ADDR”] = addr
File “/project/user/DL/anaconda3/envs/modulus2203/lib/python3.9/os.py”, line 684, in setitem
value = self.encodevalue(value)
File “/project/user/DL/anaconda3/envs/modulus2203/lib/python3.9/os.py”, line 756, in encode
raise TypeError(“str expected, not %s” % type(value).name)
TypeError: str expected, not NoneType

[prakhar_sharma]

Plz don’t use conda env. Use Python venv. I managed to run full Nvidia Modulus i.e. with tessellation library.

[yu_cheng]

I am also trying to install Modulus on my workstation. I am wondering does Modulus v22.03 support all NVIDIA GPUs including Ampere GPUs like A5000 and Turing GPUs like RTX 4000?

[prakhar_sharma]

yes I am using RTX A5000 on Modulus. There is no limitation on the GPU as long as the CUDA version of PyTorch is compatible with the GPU driver.

[ngeneva]

Hi @yu_cheng ,

Have a look at the install guide for the system/driver requirements. These newer GPUs should work fine (A5000), its older hardware that is typically the concern. I’ll make a note to get these GPUs officially tested and listed. Is there a reason for trying v22.03 versus 22.07?

[yu_cheng]

Sounds good. Sorry I didn’t mean to install v22.03. Will try the latest version.