batch_benchmark_env[ftime]_agents[ada]_model[region].batch

#!/bin/bash
#

#SBATCH --job-name=f_time
#SBATCH --output=out.txt
#SBATCH --error=out.txt

## For partition: either prod10, prod 20, prod 40 or prod80
## For gres: either 1g.10gb:[1:10] for prod10, 2g.20gb:[1:4] for prod20, 3g.40gb:1 for prod40 or A100.80gb for prod80.

##SBATCH --partition=prod10
##SBATCH --gres=gpu:1g.10gb:1
##SBATCH --cpus-per-task=4

#SBATCH --partition=prod20
#SBATCH --gres=gpu:2g.20gb:1
#SBATCH --cpus-per-task=4

##SBATCH --partition=prod40
##SBATCH --gres=gpu:3g.40gb:1
##SBATCH --cpus-per-task=4

##SBATCH --partition=prod80
##SBATCH --gres=gpu:A100.80gb:1
##SBATCH --ntasks-per-node=1
##SBATCH --cpus-per-task=8
##SBATCH --mem-per-cpu=10G
##SBATCH --nodes=1

## For ntasks and cpus: total requested cpus (ntasks * cpus-per-task) must be in [1: 4 * nMIG] with nMIG = nb_1g.10gb | 2 * nb_2g.20gb | 4 * nb_3g.40gb | 8 * nb_A100.80gb


## N tasks
#SBATCH --ntasks=1

## Walltime limit
#SBATCH --time=24:00:00

## Setup
source ~/projects/EcoJAX/venv_linux/bin/activate
cd ~/projects/EcoJAX

## Create a directory to store the logs
initial_date=$(date +"%Y%m%d_%H%M%S")

## Iterate over the seeds
seed_max=100
benchmark_name='bench5_f_time'
for _ in $(seq 1 $seed_max); do
    seed=$RANDOM
    for p_base_fruit_growth in 0.0065 0.008 0.01; do
        for variability_fruits in "[0, 0, 0, 0]" "[0, 0.1, 1, 5]" "[0, 0.2, 5, 17]" "[1, 3, 5, 9]"; do
            log_dir="logs/run_$benchmark_name-$initial_date/p_base_fruit_growth_$p_base_fruit_growth/variability_fruits_$variability_fruits"
            mkdir -p "$log_dir"
            python run.py --config-name dgx do_wandb=True env/metrics=metrics_dgx +benchmark_name=$benchmark_name seed=$seed \
                agents=ada \
                model=region \
                env=fruits \
                agents.do_include_id_fruit=True \
                env.mode_variability_fruits=time \
                env.p_base_fruit_growth=$p_base_fruit_growth \
                env.variability_fruits="$variability_fruits" \
                model.mlp_region_config.hidden_dims=[] \
                model.mlp_config.hidden_dims=[10] \
                > "$log_dir"/seed_"$seed".log 2>&1  
        done
    done
done