training_time.sh

# SESSION_NAME="dreamcatcher_train_neurips25_final1"

# # Check if the session already exists
# if tmux has-session -t $SESSION_NAME 2>/dev/null; then
#     echo "Session $SESSION_NAME already exists. Attaching to it."
#     tmux attach-session -t $SESSION_NAME
# else
#     # Create a new session and name it
#     tmux new-session -d -s $SESSION_NAME
    
seed=10
critic_choice=""
# 10 Robot 20 Task 200 Problems
critic="makespan-atari${critic_choice}"
conda activate Dreamcatcher
# run simultaneous
# declare -a MODELS=("hgt" "hgt_resnet" "hgt_edge" "hgt_edge_resnet")
# for model in "${MODELS[@]}"
# do
#     export PYTHONHASHSEED=0 # ensures full reprodibility and internal consistency
#     python training/sac_wip_simultaneous.py --env_location data/problem_set_r10_t20_s0_f10_w25_euc_2000_uni --end_problem 200 --env_id "paper aamas25 10r20t0s10f25w200p ${model} simultaneous reward-greedy-final1" --num_heads 1 --num_layers 4 --no-partition_learning --graph_mode ${model} --baseline_boosting --seed ${seed} --adaptive_temperature --gamma 0.95 --timeit final_results/training_time.csv --num_iterations 100 --no-track
# done
# run sequential
# declare -a MODELS=("hetgat" "hetgat_resnet" "hgt" "hgt_resnet" "hgt_edge" "hgt_edge_resnet")
declare -a MODELS=("hgt_resnet")
for model in "${MODELS[@]}"
do
    
    export PYTHONHASHSEED=0 # ensures full reprodibility and internal consistency
    python training/sac_wip.py --env_location data/problem_set_r10_t20_s0_f10_w25_euc_2000_uni --end_problem 200 --env_id "paper neurips25 timeit 10r20t0s10f25w200p ${model} ${critic} reward-greedy-final1" --num_heads 8 --num_layers 4 --no-partition_learning --reward_mode ${critic} --critic_mode ${model} --graph_mode ${model} --no-baseline_boosting --seed ${seed} --adaptive_temperature --timeit final_results/training_time.csv --num_iterations 100 --no-track

done