Hi,
This is not an issue, but rather a suggestion.
Recently I have been developing a tool to ease the process of hyper-parameter tuning for Deep Learning projects and I came up with GridSearcher. Besides the multi-gpu functionality and the basic resource management, it also has a class to run experiments with SBATCH. It allows adding environment variables and command line arguments for sbatch using python dictionaries. I would suggest using it in scripts/slurm_sbatcher.py because it would simplify the code by a lot. Copying the example and replacing the existing code should be easy to do.
You can install it via pip install gridsearcher and you can find more information here:
I hope it will be useful.
Thank you,
Ionut
Hi,
This is not an issue, but rather a suggestion.
Recently I have been developing a tool to ease the process of hyper-parameter tuning for Deep Learning projects and I came up with
GridSearcher. Besides the multi-gpu functionality and the basic resource management, it also has a class to run experiments with SBATCH. It allows adding environment variables and command line arguments forsbatchusing python dictionaries. I would suggest using it inscripts/slurm_sbatcher.pybecause it would simplify the code by a lot. Copying the example and replacing the existing code should be easy to do.You can install it via
pip install gridsearcherand you can find more information here:I hope it will be useful.
Thank you,
Ionut