[SE(3)-Transformer/All] Minor issue in default parameter for logging.: Change default from "/results" to "./results".

[RMichae1]

Related to SE(3)-Transformer/All

Describe the bug The default log_dir argument is set to "/results" instead of referencing the current/working directory "./results", which fails when running the scripts/train.sh with the provided default arguments. This results in PermissionError: [Errno 13] Permission denied: '/results' (- when working on a Linux system). The root-cause can be found in the runtime/arguments.py default argument: https://github.com/NVIDIA/DeepLearningExamples/blob/35d8759cb8cf52f8c7d33900ef27fd0f16d6cff3/DGLPyTorch/DrugDiscovery/SE3Transformer/se3_transformer/runtime/arguments.py#L36

To Reproduce Steps to reproduce the behavior:

1. Setup of SE(3)-Transformer either via provided Dockerfile or reproducing the environment as specified .
2. Run 'bash ./scripts/train.sh' with default arguments

Expected behavior The script train.sh should not terminate with an error and should log output to the default specified directory that is not "/results", but on relative path of execution "./results" .

Environment Please provide at least:

• Container version (e.g. pytorch:19.05-py3): nvcr.io/nvidia/pytorch:22.08-py3
• GPUs in the system: (e.g. 8x Tesla V100-SXM2-16GB): 1x A100-40GB
• CUDA driver version (e.g. 418.67): 525.50.13

[milesial]

Hi, I assume this only appears when running outside of docker right? When using the recommended docker image, this shouldn't appear since the docker user is root. I'll get this changed anyway if it makes life easier for non-docker users.

[RMichae1]

Hi, You're correct! This was run outside of Docker. Thank you for addressing it.

[milesial]

Fix landed in https://github.com/NVIDIA/DeepLearningExamples/commit/2586ee38a7c3ec74cb7d565558e72175334511c7