We run our RL algorithm on any Minihack environments (RoomRandom-15x15 or RoomTrap-15x15). Occasionally after around 1M steps we obtain the following FileNotFoundError when looking for the NetHack tmp directory:
Expected behavior
No FileNotFoundError when running our RL algorithm on MiniHack.
Environment
I'm running this on a SLURM scheduler so didn't collect the GPU info.
Collecting environment information...
MiniHack version: 0.1.3
NLE version: 0.8.1
Gym version: 0.23.1
PyTorch version: N/A
Is debug build: N/A
CUDA used to build PyTorch: N/A
OS: CentOS Linux release 8.1.1911 (Core)
GCC version: (conda-forge gcc 10.3.0-16) 10.3.0
CMake version: version 3.23.2
Python version: 3.8
Is CUDA available: N/A
CUDA runtime version: Could not collect
GPU models and configuration: Could not collect
Nvidia driver version: Could not collect
cuDNN version: Could not collect
Versions of relevant libraries:
[pip3] numpy==1.19.5
[conda] Could not collect
This looks like an NLE problem as the directory creation goes back to there. That, and/or a problem with your computer deleting temp files automatically after a certain amount of time?
🐛 Bug
We get a file not found error when looking for the NetHack temporary directory after running our RL algorithm for > 1M steps. This error doesn't occur on all runs, only some runs get this error. We use the following environment wrapper over MiniHack: https://github.com/AGI-Labs/continual_rl/blob/develop/continual_rl/experiments/tasks/make_minihack_task.py.
To Reproduce
Steps to reproduce the behavior:
We run our RL algorithm on any Minihack environments (RoomRandom-15x15 or RoomTrap-15x15). Occasionally after around 1M steps we obtain the following FileNotFoundError when looking for the NetHack tmp directory:
Expected behavior
No FileNotFoundError when running our RL algorithm on MiniHack.
Environment
I'm running this on a SLURM scheduler so didn't collect the GPU info.
Collecting environment information... MiniHack version: 0.1.3 NLE version: 0.8.1 Gym version: 0.23.1 PyTorch version: N/A Is debug build: N/A CUDA used to build PyTorch: N/A
OS: CentOS Linux release 8.1.1911 (Core) GCC version: (conda-forge gcc 10.3.0-16) 10.3.0 CMake version: version 3.23.2
Python version: 3.8 Is CUDA available: N/A CUDA runtime version: Could not collect GPU models and configuration: Could not collect Nvidia driver version: Could not collect cuDNN version: Could not collect
Versions of relevant libraries: [pip3] numpy==1.19.5 [conda] Could not collect
Additional context
-