#################
actions = {}
## Set high episode to test convergence
# Change back to resonable setting for other testing
n_episodes = 2500
#########################
# Statistic variables #
#########################
k = 50 # Used for average win rates
p = 5 # Print episodic results every p episodes
stats = AgentStatistics(names[0],
n_episodes,
k,
save_file=os.getcwd() +
"/saved-stats/A2C_test_2_stats")
scores = []
short_term_wr = np.zeros((k, ), dtype=int) # Used to average win rates
short_term_scores = [0.5] # Average win rates per k episodes
ties = 0
losses = 0
score = 0
current_eps = 0
epsilonVals = []
current_loss = 0
lossVals = []
average_reward = 0
avgRewardVals = []
players[1] = random_actions(env.num_actions_per_turn, 1, map_name)
names[1] = 'Random Agent'
#################
actions = {}
## Set high episode to test convergence
# Change back to resonable setting for other testing
n_episodes = 20000
#########################
# Statistic variables #
#########################
k = 100
stats = AgentStatistics(names[0],
n_episodes,
k,
save_file='./saved-stats/smart_state_newton')
short_term_wr = np.zeros((k, ), dtype=int) # Used to average win rates
ties = 0
losses = 0
score = 0
current_eps = 0
current_loss = 0
q_values = 0
reward = {0: 0, 1: 0}
#####################
# Training Loop #
players[1] = base_rushV1(env.num_actions_per_turn, 1)
names[1] = 'Base Rush v1'
#################
actions = {}
## Set high episode to test convergence
# Change back to resonable setting for other testing
n_episodes = 50
#########################
# Statistic variables #
#########################
k = 100
stats = AgentStatistics(names[0],
n_episodes,
k,
save_file='/saved-stats/local')
short_term_wr = np.zeros((k, ), dtype=int) # Used to average win rates
ties = 0
losses = 0
score = 0
current_eps = 0
current_loss = 0
q_values = 0
reward = {0: 0, 1: 0}
#####################
# Training Loop #
actions = {}
## Set high episode to test convergence
# Change back to resonable setting for other testing
n_episodes = 2500
RENDER_CHARTS = True # Determines if final charts should be rendered
timestep = 0
#########################
# Statistic variables #
#########################
k = 100 #The set number of episodes to show win rates for
# The Stats class (for saving statistics)
stats = AgentStatistics(names[0],
n_episodes,
k,
save_file="/saved-stats/rppo_newton_v14")
# General stats
score = 0
losses = 0
ties = 0
# Short wr
short_term_wr = np.zeros((k, ), dtype=int) # Used to average win rates
# Epsilon and losses
current_eps = 0
current_loss = 0
current_actor_loss = 0
current_critic_loss = 0
from utils.Statistics import AgentStatistics from agents.Smart_State.render_smart_state import render_charts # Create and load the statistics SAVED_STATS_PATH = 'saved-stats/best_smart_state' stats = AgentStatistics() stats.load_stats(SAVED_STATS_PATH) # Render the charts render_charts(stats)
names[0] = "DQN Agent"
players[1] = random_actions_delay(env.num_actions_per_turn, 1, map_name)
names[1] = 'Random Agent Delay'
#################
actions = {}
## Set high episode to test convergence
# Change back to resonable setting for other testing
n_episodes = 5000
#########################
# Statistic variables #
#########################
k = 100
stats = AgentStatistics(names[0], n_episodes, k, save_file="saved-stats/dqn_new")
short_term_wr = np.zeros((k,), dtype=int) # Used to average win rates
ties = 0
losses = 0
score = 0
current_eps = 0
current_loss = 0
q_values = 0
reward = {0: 0, 1: 0}
#########################
#####################
# Training Loop #