update_resign_threshold.py

# Copyright 2018 Google LLC
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#      http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

"""Updates a flagfile containing a resign threshold flag

Reads the bigtables defined by the flags --cbt_{project, instance, table} to
compute the 95 percentile of the bleakest-evaluations found in calibration
games, then updates the flagfile on the default bucket path, resetting that
value.

Recommended usage is via common flagfile (e.g. rl_loop/distributed_flags)
"""


import sys
import re
import os
import time
from absl import app
from absl import flags
import numpy as np
import tensorflow as tf

sys.path.insert(0, '.')

import mask_flags
import bigtable_input
import rl_loop.fsdb as fsdb

# Fun fact, this only helps for --helpshort.  It's not a validator.
flags.adopt_module_key_flags(bigtable_input)
flags.adopt_module_key_flags(fsdb)

FLAGS = flags.FLAGS
RESIGN_FLAG_REGEX = re.compile(r'--resign_threshold=([-\d.]+)')

def get_95_percentile_bleak(games_nr, n_back=500):
    """Gets the 95th percentile of bleakest_eval from bigtable"""
    end_game = int(games_nr.latest_game_number)
    start_game = end_game - n_back if end_game >= n_back else 0
    moves = games_nr.bleakest_moves(start_game, end_game)
    evals = np.array([m[2] for m in moves])
    return np.percentile(evals, 5)

def update_flagfile(flags_path, new_threshold):
    """Updates the flagfile at `flags_path`, changing the value for
    `resign_threshold` to `new_threshold`
    """
    if abs(new_threshold) > 1:
        raise ValueError("Invalid new percentile for resign threshold")
    with tf.gfile.GFile(flags_path) as f:
        lines = f.read()
    if new_threshold > 0:
        new_threshold *= -1
    if not RESIGN_FLAG_REGEX.search(lines):
        print("Resign threshold flag not found in flagfile {}!  Aborting.".format(flags_path))
        sys.exit(1)
    old_threshold = RESIGN_FLAG_REGEX.search(lines).groups(1)
    lines = re.sub(RESIGN_FLAG_REGEX, "--resign_threshold={:.3f}".format(new_threshold), lines)

    if abs(float(old_threshold[0]) - new_threshold) < 0.001:
        print("Not updating percentiles; {} ~= {:.3f}".format(
                old_threshold[0], new_threshold), flush=True)
    else:
        print("Updated percentile from {} to {:.3f}".format(
                old_threshold[0], new_threshold), flush=True)
        with tf.gfile.GFile(flags_path, 'w') as f:
            f.write(lines)


def main(argv):
    if len(argv) > 1:
        raise app.UsageError('Too many command-line arguments.')

    games_nr = bigtable_input.GameQueue(
        FLAGS.cbt_project, FLAGS.cbt_instance, FLAGS.cbt_table + '-nr')

    while True:
        new_pct = get_95_percentile_bleak(games_nr)
        update_flagfile(fsdb.flags_path(), new_pct)
        time.sleep(60 * 3)

if __name__ == '__main__':
    valid_flags = list(map(lambda f: '--' + f, FLAGS.flag_values_dict().keys()))
    valid_flags += ['--helpshort', '--helpfull', '--help']
    parsed_flags = flags.FlagValues().read_flags_from_files(sys.argv[1:])
    filtered_flags = mask_flags.filter_flags(parsed_flags, valid_flags)
    print(filtered_flags, flush=True)
    app.run(main, argv=sys.argv[:1] + filtered_flags)