ikbendewilliam

NUMBER_OF_GAMES = 5000
begintime = time.time()
for game in range(NUMBER_OF_GAMES):
    board_log, action_log, probs_log, result = play_game()
    all_boards.append(board_log[-1][-1])
    all_actions.append(action_log)
    all_probs.append(probs_log)
    if (game + 1) % math.floor(NUMBER_OF_GAMES / 100) == 0 and game > 0:
        time_passed = math.floor(time.time() - begintime)
        print(str(game) + " / " + str(NUMBER_OF_GAMES) + \

ikbendewilliam

botName = 'yourbotname'

from random import randint
from random import choice
import tensorflow as tf
import numpy as np
import time
import datetime
import math
import requests

ikbendewilliam

def benchmark(games = 100):
    wins = draws = losses = 0
    for i in range(games):
        player_index = 0
        player_ids = ["p0", "p1"]
        result = initialise(player_index, player_ids, BOARD_SIZE)

        # Initialise logs for game
        board_log = []
        action_log = [[], []]

ikbendewilliam

print(benchmark(1000))

ikbendewilliam

def guess_move(current_board, possible_moves, training):
    probs = session.run(probabilities, feed_dict={input_positions:[[cell for row in current_board for cell in row]]})[0]

    possible_moves = [move[0] * BOARD_SIZE[1] + move[1] for move in possible_moves]
    probs = [p if index in possible_moves else 0 for index, p in enumerate(probs)]
    probs = [round(p, 6) for p in probs]

    if sum(probs) > 0:
        probs = [p / sum(probs) for p in probs]
    else:

ikbendewilliam

import aigamingReversingStones
import tensorflow as tf
import numpy as np
import time
import datetime
import math
from random import randint
from random import choice

ikbendewilliam

CONFIGURATIONS = [[8, 8], [14, 8], [4, 4]]
USE_CONFIGURATION = 0

BOARD_SIZE = CONFIGURATIONS[USE_CONFIGURATION]
print("Using configuration: \nBoard size: " + str(BOARD_SIZE[0]) + " by " + str(BOARD_SIZE[1]))

ikbendewilliam

inputs_units = BOARD_SIZE[0] * BOARD_SIZE[1]
hidden_units = BOARD_SIZE[0] * BOARD_SIZE[1]
output_units = BOARD_SIZE[0] * BOARD_SIZE[1]

def initialise_tf():
    global input_positions, labels, learning_rate, W1, b1, h1, W2, b2, logits, probabilities, cross_entropy, train_step
    input_positions = tf.placeholder(tf.float32, shape=(1, inputs_units))
    labels          = tf.placeholder(tf.int64)
    learning_rate   = tf.placeholder(tf.float32, shape=[])

ikbendewilliam

def initialise(player_index, player_ids, board_dimensions):
    return aigamingReversingStones.initialise(player_index=player_index, player_ids=player_ids, board_dimensions=board_dimensions)

ikbendewilliam

def load_model(name='model.ckpt', games = 100):
    global session
    tf.reset_default_graph()
    initialise_tf()
    
    session = tf.Session()
    tf.train.Saver().restore(session, "./models/" + name)
    benchmark(games)