import sys
import os
sys.path.append(os.path.join(os.path.dirname(__file__), os.path.pardir))
from torch.utils.tensorboard import SummaryWriter
from src.env.grid_env import GridEnv
from src.agent import Agent
import numpy as np
import pygame
visualise = True
grid_env = GridEnv(visualize=visualise,
field_file_path="src/data/ori_data.txt",
max_step=500)
BATCH_SIZE = 256
GAMMA = 0.8
EPS_START = 0.5
EPS_END = 0.8
EPS_DECAY = 200
TARGET_UPDATE = 10
num_episodes = 5000000
best = dict()
dqn_agent = Agent()
writer = SummaryWriter(log_dir=os.path.join("log"))
# output
'output_folder': "output_actor_critic",
'log_folder': 'log',
'model_folder': 'model',
'memory_config_dir': "memory_config"
}
params['log_folder'] = os.path.join(params['output_folder'], params['log_folder'])
params['model_folder'] = os.path.join(params['output_folder'], params['model_folder'])
if not os.path.exists(params['log_folder']):
os.makedirs(params['log_folder'])
if not os.path.exists(params['model_folder']):
os.makedirs(params['model_folder'])
painter = Painter(params) if params['visualise'] else None
grid_env = GridEnv(params, painter)
model_path = os.path.join(params['output_folder'], "model", "Agent_dqn_state_dict_1600.mdl")
agent_ac = Agent(params, painter)
writer = SummaryWriter(log_dir=params['log_folder'])
all_mean_rewards = []
all_mean_actor_loss = []
all_mean_critic_loss = []
time_step = 0
for i_episode in range(params['num_episodes']):
observed_map, robot_pose = grid_env.reset()
done = False
rewards = []
actor_losses = []
'model_folder': 'model',
'memory_config_dir': "memory_config",
'use_cuda': True
}
params['log_folder'] = os.path.join(params['output_folder'],
params['log_folder'])
params['model_folder'] = os.path.join(params['output_folder'],
params['model_folder'])
if not os.path.exists(params['log_folder']):
os.makedirs(params['log_folder'])
if not os.path.exists(params['model_folder']):
os.makedirs(params['model_folder'])
painter = Painter(params) if params['visualise'] else None
grid_env = GridEnv(params, painter)
train_device = 'cuda' if torch.cuda.is_available(
) and params['use_cuda'] else 'cpu'
writer = SummaryWriter(log_dir=params['log_folder'])
player = Agent(params=params,
writer=writer,
train_agent=True,
is_resume=False,
filepath=None,
train_device=torch.device(train_device))
total_rewards, smoothed_rewards = [], []
global_step = 0