search

stark-t / PAI

Pollination_Artificial_Intelligence
5 stars 1 forks source link

Simplify the cluster scripts by reducing to a single general slurm script #69

Closed valentinitnelav closed 1 year ago

valentinitnelav commented 1 year ago

Right now I had several slurm scripts for each type of model. It felt easier like that when I started the experimental work on the cluster, but now I can write a general slurm script that is covering all the particular cases and it can be called with parameters for each type of model (be it YOLOv5 or YOLOv7).

So instead of having multiple *train* and *detect* Slurm scripts, I can have just two scripts train.sh and detect.sh that can be executed from the server's terminal like:

sbatch <options> /path/to/train.sh \
    data_folder \
    model_name \
    job_weights \
    job_hyp \
    job_epochs \
    job_batch_size \
    job_img_size \
    job_workers
valentinitnelav commented 1 year ago

Note that this is optional. The repo get be published without this beautification as well

valentinitnelav commented 1 year ago

I will not have extra time for this. This is good enough as such.