def main(stdscr, model_path: str, num: int, result_logger: ResultLogger):
terminal = Terminal(stdscr, create_char_map())
terminal.init_window(FIELD_WIDTH, FIELD_ROW)
program_set = BubblesortProgramSet()
Bubblesort_env = BubblesortEnv(FIELD_ROW, FIELD_WIDTH, FIELD_DEPTH)
questions = create_questions(num)
if DEBUG_MODE:
questions = questions[-num:]
system = RuntimeSystem(terminal=terminal)
npi_model = BubblesortNPIModel(system, model_path, program_set)
npi_runner = TerminalNPIRunner(terminal, npi_model, recording=False)
npi_runner.verbose = DEBUG_MODE
correct_count = wrong_count = 0
for data in questions:
Bubblesort_env.reset()
try:
run_npi(Bubblesort_env, npi_runner, program_set.BUBBLESORT, data)
if data['correct']:
correct_count += 1
else:
wrong_count += 1
except StopIteration:
wrong_count += 1
pass
result_logger.write(data)
terminal.add_log(data)
return correct_count, wrong_count
def main(stdscr, model_path: str, num: int, result_logger: ResultLogger):
terminal = Terminal(stdscr, create_char_map())
terminal.init_window(FIELD_WIDTH, FIELD_ROW)
program_set = MultiplicationProgramSet()
multiplyition_env = MultiplicationEnv(FIELD_ROW, FIELD_WIDTH, FIELD_DEPTH)
questions = create_questions(num)
if DEBUG_MODE:
questions = questions[-num:]
system = RuntimeSystem(terminal=terminal)
npi_model = MultiplicationNPIModel(system, terminal, model_path,
program_set)
npi_runner = TerminalNPIRunner(terminal, npi_model, recording=False)
npi_runner.verbose = DEBUG_MODE
correct_count = wrong_count = 0
for data in questions:
multiplyition_env.reset()
run_npi(multiplyition_env, npi_runner, program_set.ADD, data)
result_logger.write(data)
terminal.multiply_log(data)
if data['correct']:
correct_count += 1
else:
wrong_count += 1
return correct_count, wrong_count
def main(stdscr, filename: str, num_of_questions: int,
result_logger: ResultLogger):
terminal = Terminal(stdscr, create_char_map())
terminal.init_window(FIELD_WIDTH, FIELD_ROW)
program_set = MultiplicationProgramSet()
multiplication_env = MultiplicationEnv(FIELD_ROW, FIELD_WIDTH, FIELD_DEPTH,
terminal)
questions = create_questions(num_of_questions)
teacher = MultiplicationTeacher(program_set, terminal)
npi_runner = TerminalNPIRunner(terminal, teacher, result_logger)
npi_runner.verbose = DEBUG_MODE
steps_list = []
for data in questions:
multiplication_env.reset()
q = copy(data)
run_npi(multiplication_env, npi_runner, program_set.MUL, data)
steps_list.append({"q": q, "steps": npi_runner.step_list})
result_logger.write(data)
terminal.add_log(data)
if filename:
with open(filename, 'wb') as f:
pickle.dump(steps_list, f, protocol=pickle.HIGHEST_PROTOCOL)
def main(stdscr, model_path: str, num: int, result_logger: ResultLogger):
terminal = Terminal(stdscr, create_char_map())
terminal.init_window(FIELD_WIDTH, FIELD_ROW)
program_set = BubblesortProgramSet()
Bubblesort_env = BubblesortEnv(FIELD_ROW, FIELD_WIDTH, FIELD_DEPTH)
questions = create_questions(num)
if DEBUG_MODE:
questions = questions[-num:]
system = RuntimeSystem(terminal=terminal)
npi_model = BubblesortNPIModel(system, model_path, program_set)
npi_runner = TerminalNPIRunner(terminal, npi_model, recording=False)
npi_runner.verbose = DEBUG_MODE
correct_count = wrong_count = 0
slot_num = 10 # 10 means 10 slots, each
# slot's size is 0.1
for i in range(slot_num):
arr.append([])
# Put array elements in different buckets
for j in x:
index_b = int(slot_num * j)
arr[index_b].append(j)
# Sort individual buckets
for i in range(slot_num):
Bubblesort_env.reset()
try:
run_npi(Bubblesort_env, npi_runner, program_set.BUBBLESORT, data)
if data['correct']:
correct_count += 1
else:
wrong_count += 1
except StopIteration:
wrong_count += 1
pass
result_logger.write(data)
terminal.add_log(data)
return correct_count, wrong_count
# concatenate the result
k = 0
for i in range(slot_num):
for j in range(len(arr[i])):
x[k] = arr[i][j]
k += 1
return x
def main(stdscr, filename: str, num: int, result_logger: ResultLogger):
terminal = Terminal(stdscr, create_char_map())
terminal.init_window(FIELD_WIDTH, FIELD_ROW)
program_set = AdditionProgramSet()
addition_env = AdditionEnv(FIELD_ROW, FIELD_WIDTH, FIELD_DEPTH)
questions = create_questions(num)
teacher = AdditionTeacher(program_set)
npi_runner = TerminalNPIRunner(terminal, teacher)
npi_runner.verbose = DEBUG_MODE
steps_list = []
for data in questions:
addition_env.reset()
q = copy(data)
run_npi(addition_env, npi_runner, program_set.ADD, data)
steps_list.append({"q": q, "steps": npi_runner.step_list})
result_logger.write(data)
terminal.add_log(data)
if filename:
with open(filename, 'wb') as f:
pickle.dump(steps_list, f, protocol=pickle.HIGHEST_PROTOCOL)
def main(stdscr, filename: str, num: int, result_logger: ResultLogger):
terminal = Terminal(stdscr, create_char_map())
terminal.init_window(FIELD_WIDTH, FIELD_ROW)
program_set = BubblesortProgramSet()
bubblesort_env = BubblesortEnv(FIELD_ROW, FIELD_WIDTH, FIELD_DEPTH)
questions = create_questions(number=num)
teacher = BubblesortTeacher(program_set)
npi_runner = TerminalNPIRunner(terminal, teacher)
npi_runner.verbose = DEBUG_MODE
steps_list = []
f = open("debug_log.csv", "w")
for data in questions:
bubblesort_env.reset()
q = copy(data)
run_npi(bubblesort_env, npi_runner, program_set.BUBBLESORT, data)
steps_list.append({"q": q, "steps": npi_runner.step_list})
for step in npi_runner.step_list:
f.write("{},".format(step.input.env[0]))
f.write("{},".format(step.input.env[1]))
f.write("{},".format(step.input.env[2]))
f.write("{},".format(step.input.env[3]))
f.write("{},".format(step.input.env[4]))
f.write("{},".format(step.input.env[5]))
f.write("{},".format(step.input.arguments))
f.write("{},".format(step.input.program))
f.write("{},".format(step.output.program))
f.write("{},".format(step.output.r))
f.write("{},".format(step.output.arguments))
f.write("{}\n".format(q))
result_logger.write(data)
terminal.add_log(data)
f.close()
if filename:
with open(filename, 'wb') as f:
pickle.dump(steps_list, f, protocol=pickle.HIGHEST_PROTOCOL)
def main(stdscr, model_path: str, num: int, result_logger: ResultLogger):
terminal = Terminal(stdscr, create_char_map())
terminal.init_window(FIELD_WIDTH, FIELD_ROW)
program_set = AdditionProgramSet()
addition_env = AdditionEnv(FIELD_ROW, FIELD_WIDTH, FIELD_DEPTH)
questions = create_questions(num)
if DEBUG_MODE:
questions = questions[-num:]
system = RuntimeSystem(terminal=terminal)
npi_model = AdditionNPIModel(system, model_path, program_set)
npi_runner = TerminalNPIRunner(terminal, npi_model, recording=False)
npi_runner.verbose = DEBUG_MODE
correct_count = wrong_count = 0
for data in questions:
addition_env.reset()
run_npi(addition_env, npi_runner, program_set.ADD, data)
result_logger.write(data)
terminal.add_log(data)
if data["correct"]:
correct_count += 1
else:
wrong_count += 1
return correct_count, wrong_count
def __init__(self,
terminal: Terminal,
model: NPIStep = None,
recording=True,
max_depth=100,
max_step=10000,
result_logger=ResultLogger('result_multiplication.log')):
self.terminal = terminal
self.model = model
self.steps = 0
self.step_list = []
self.alpha = 0.5
self.verbose = True
self.recording = recording
self.max_depth = max_depth
self.max_step = max_step
self.result_logger = result_logger
questions = create_questions(num)
teacher = AdditionTeacher(program_set)
npi_runner = TerminalNPIRunner(terminal, teacher)
npi_runner.verbose = DEBUG_MODE
steps_list = []
for data in questions:
addition_env.reset()
q = copy(data)
run_npi(addition_env, npi_runner, program_set.ADD, data)
steps_list.append({"q": q, "steps": npi_runner.step_list})
result_logger.write(data)
terminal.add_log(data)
if filename:
with open(filename, 'wb') as f:
pickle.dump(steps_list, f, protocol=pickle.HIGHEST_PROTOCOL)
if __name__ == '__main__':
import sys
DEBUG_MODE = os.environ.get('DEBUG')
if DEBUG_MODE:
output_filename = None
num_data = 3
log_filename = 'result.log'
else:
output_filename = sys.argv[1] if len(sys.argv) > 1 else None
num_data = int(sys.argv[2]) if len(sys.argv) > 2 else 1000
log_filename = sys.argv[3] if len(sys.argv) > 3 else 'result.log'
curses.wrapper(main, output_filename, num_data, ResultLogger(log_filename))
print("create %d training data" % num_data)
system = RuntimeSystem(terminal=terminal)
npi_model = BubblesortNPIModel(system, model_path, program_set)
npi_runner = TerminalNPIRunner(terminal, npi_model, recording=False)
npi_runner.verbose = DEBUG_MODE
correct_count = wrong_count = 0
for data in questions:
Bubblesort_env.reset()
try:
run_npi(Bubblesort_env, npi_runner, program_set.BUBBLESORT, data)
if data['correct']:
correct_count += 1
else:
wrong_count += 1
except StopIteration:
wrong_count += 1
pass
result_logger.write(data)
terminal.add_log(data)
return correct_count, wrong_count
if __name__ == '__main__':
import sys
DEBUG_MODE = os.environ.get('DEBUG')
model_path_ = sys.argv[1]
num_data = int(sys.argv[2]) if len(sys.argv) > 2 else 1000
log_filename = sys.argv[3] if len(sys.argv) > 3 else 'result.log'
cc, wc = curses.wrapper(main, model_path_, num_data,
ResultLogger(log_filename))
print("Accuracy %s(OK=%d, NG=%d)" % (cc / (cc + wc), cc, wc))