for scenario in scenarios:
for seed in seeds:
command = [
"python", "ppo_main_nlp.py",
"--scenario",scenario,
"--rep-type","nlp",
"--seed",str(seed),
"--env-name","doom",
"--button_number",str(button_number),
"--algo","ppo",
"--lr",str(lr),
"--value-loss-coef",str(value_loss_coef),
"--num-processes",str(num_processes),
"--num-env-steps",str(num_env_steps),
"--num-steps",str(num_steps),
"--num-mini-batch",str(num_mini_batch),
"--log-interval",str(1),
"--entropy-coef",str(0.01),
"--n-channels",str(1),
"--n-patches",str(5)
]
p = subprocess.Popen(command,shell=False)
processes.append(p)
for p in processes:
p.wait()
for scenario in scenarios:
final_plot(scenario, num_updates, len(seeds))
"defend_the_center_extreme", "health_gathering_extreme"
]
seeds = [35, 45, 59, 12, 5]
processes = []
steps_per_epoch = 500
epochs = 100
frame_repeat = 4
for scenario in scenarios:
for seed in seeds:
command = [
"python", "dqn_nlp.py", "--SCENARIO", scenario, "--REP_TYPE",
"nlp", "--SEED",
str(seed), "--BATCH_SIZE", "100", "--ARCH", "TextCNN",
"--SENTANCE_LEN", "200", "--LEARNING_STEPS_PER_EPOCH",
str(steps_per_epoch), "--HIDDEN_UNITS", "16", "--FILTER_COUNT",
"12", "--LEARNING_RATE", "0.00025", "--EPOCHS",
str(epochs), "--FRAME_REPEAT",
str(frame_repeat)
]
p = subprocess.Popen(command, shell=False)
processes.append(p)
for p in processes:
p.wait()
for scenario in scenarios:
final_plot(scenario, steps_per_epoch, len(seeds))