-
Notifications
You must be signed in to change notification settings - Fork 5
Help
fsmosca edited this page Mar 11, 2021
·
9 revisions
usage: Lakas v0.35.0 [-h] --engine ENGINE [--protocol PROTOCOL] [--base-time-sec BASE_TIME_SEC] [--inc-time-sec INC_TIME_SEC] [--depth DEPTH] [--move-time-ms MOVE_TIME_MS] [--nodes NODES] [--optimizer OPTIMIZER]
[--oneplusone-noise-handling ONEPLUSONE_NOISE_HANDLING] [--oneplusone-mutation ONEPLUSONE_MUTATION] [--oneplusone-crossover ONEPLUSONE_CROSSOVER] [--tbpsa-naive TBPSA_NAIVE]
[--tbpsa-initial-popsize TBPSA_INITIAL_POPSIZE] [--bo-utility-kind BO_UTILITY_KIND] [--bo-utility-kappa BO_UTILITY_KAPPA] [--bo-utility-xi BO_UTILITY_XI] [--bo-initialization BO_INITIALIZATION]
[--bo-gp-param-alpha BO_GP_PARAM_ALPHA] [--spsa-scale SPSA_SCALE] [--budget BUDGET] [--concurrency CONCURRENCY] [--games-per-budget GAMES_PER_BUDGET] [--match-manager MATCH_MANAGER] --match-manager-path
MATCH_MANAGER_PATH --opening-file OPENING_FILE --opening-file-format OPENING_FILE_FORMAT [--variant VARIANT] [--input-data-file INPUT_DATA_FILE] [--output-data-file OUTPUT_DATA_FILE]
[--optimizer-log-file OPTIMIZER_LOG_FILE] --input-param INPUT_PARAM [--common-param COMMON_PARAM] [--deterministic-function] [--use-best-param] [--best-result-threshold BEST_RESULT_THRESHOLD]
[--cutechess-debug] [--cutechess-wait CUTECHESS_WAIT]
Parameter optimizer using nevergrad library.
optional arguments:
-h, --help show this help message and exit
--engine ENGINE Engine filename or engine path and filename.
--protocol PROTOCOL Engine filename or engine path and filename, default=uci
--base-time-sec BASE_TIME_SEC
Base time in sec for time control. If depth is not defined this option should have a value.
--inc-time-sec INC_TIME_SEC
Increment time in sec for time control.
--depth DEPTH The maximum search depth that the engine is allowed.
Example:
--depth 6 ...
--move-time-ms MOVE_TIME_MS
The maximum search time in milliseconds. Example
--move-time-ms 1000
and engine is set to search at 1s. The cutechess
timemargin is set at 50ms.
--nodes NODES The maximum nodes that the engine is allowed to search. Do not use other
move control like --base-time-sec or --depth or --move-time-sec example:
--nodes 500 ...
--optimizer OPTIMIZER
Type of optimizer to use, can be oneplusone or tbpsa or bayesopt, or spsa, or cmaes, or ngopt, default=oneplusone.
--oneplusone-noise-handling ONEPLUSONE_NOISE_HANDLING
Parameter for oneplusone optimizer, can be optimistic or random,
or a tuple, default=optimistic.
Example:
--oneplusone-noise-handling random ...
--oneplusone-noise-handling optimistic ...
--oneplusone-noise-handling "('optimistic', 0.01)" ...
where:
0.01 is the coefficient (the regularity of reevaluations),
default coefficient is 0.05.
--oneplusone-mutation ONEPLUSONE_MUTATION
Parameter for oneplusone optimizer, can be gaussian or cauchy,
or discrete or discreteBSO or fastga or doublefastga or
portfolio, default=gaussian.
--oneplusone-crossover ONEPLUSONE_CROSSOVER
Parameter for oneplusone optimizer. Whether to add a genetic crossover step
every other iteration, default=false.
--tbpsa-naive TBPSA_NAIVE
Parameter for tbpsa optimizer, set to false for noisy problem, so that the best points
will be an average of the final population, default=false.
Example:
--optimizer tbpsa --tbpsa-naive true ...
--tbpsa-initial-popsize TBPSA_INITIAL_POPSIZE
Parameter for tbpsa optimizer. Initial population size, default=4xdimension.
Example:
--optimizer tbpsa --tbpsa-initial-popsize 8 ...
--bo-utility-kind BO_UTILITY_KIND
Parameter for bo optimizer. Type of utility function to use among ucb, ei and poi, default=ucb.
Example:
--optimizer bayesopt --bo-utility-kind ei ...
--bo-utility-kappa BO_UTILITY_KAPPA
Parameter for bayesopt optimizer. Kappa parameter for the utility function, default=2.576.
Example:
--optimizer bayesopt --bo-utility-kappa 2.0 ...
--bo-utility-xi BO_UTILITY_XI
Parameter for bayesopt optimizer. Xi parameter for the utility function, default=0.0.
Example:
--optimizer bayesopt --bo-utility-xi 0.01 ...
--bo-initialization BO_INITIALIZATION
Parameter for bayesopt optimizer. Can be Hammersley or random or LHS, default=Hammersley.
Example:
--optimizer bayesopt --bo-initialization random ...
--bo-gp-param-alpha BO_GP_PARAM_ALPHA
Parameter for bayesopt optimizer on gaussian process regressor, default=0.001.
Example:
--optimizer bayesopt --bo-gp-param-alpha 0.05 ...
--spsa-scale SPSA_SCALE
Parameter for spsa optimizer to increase/decrease param perturbation, default=500000.
Example:
--optimizer spsa --spsa-scale 600000 ...
--budget BUDGET Iterations to execute, default=1000.
--concurrency CONCURRENCY
Number of game matches to run concurrently, default=1.
--games-per-budget GAMES_PER_BUDGET
Number of games per iteration, default=100.
This should be even number.
--match-manager MATCH_MANAGER
Match manager name, can be cutechess or duel, default=cutechess.
--match-manager-path MATCH_MANAGER_PATH
Match manager path and/or filename. Example:
cutechess:
--match-manager-path c:/chess/tourney_manager/cutechess/cutechess-cli.exe
duel.py for xboard engines:
--match-manager-path python c:/chess/tourney_manager/duel/duel.py
or
--match-manager-path c:/python3/python c:/chess/tourney_manager/duel/duel.py
--opening-file OPENING_FILE
Start opening filename in pgn or epd format.
--opening-file-format OPENING_FILE_FORMAT
Opening format can be epd or pgn.
--variant VARIANT Game variant, default=normal
--input-data-file INPUT_DATA_FILE
Load the saved data to continue the optimization.
--output-data-file OUTPUT_DATA_FILE
Save optimization data to this file.
--optimizer-log-file OPTIMIZER_LOG_FILE
The filename of the log of certain optimization session. This file can be used to create a plot. Default=log_nevergrad.txt, Mode=append.
--input-param INPUT_PARAM
The parameters that will be optimized.
Example 1 with 1 parameter:
--input-param "{'pawn': {'init': 92, 'lower': 90, 'upper': 120}}"
Example 2 with 2 parameters:
--input-param "{'pawn': {'init': 92, 'lower': 90, 'upper': 120}}, 'knight': {'init': 300, 'lower': 250, 'upper': 350}}"
--common-param COMMON_PARAM
The parameters that will be sent to both test and base engines.
Make sure that this param is not included in the input-param.
Example:
--common-param "{'RookOpenFile': 92, 'KnightOutpost': 300}"
--deterministic-function
A flag to consider the objective function as deterministic.
--use-best-param Use best param for the base engine. A param becomes best if it defeats the
current best by --best-result-threshold value.
--best-result-threshold BEST_RESULT_THRESHOLD
When match result is greater than this, update the best param, default=0.5.
Only applied when the flag --use-best-param is enabled, the best param will be used by the
base engine against the test engine that uses the param from the optimizer.
--cutechess-debug Enable -debug flag of cutechess-cli, this will output engine logging.
--cutechess-wait CUTECHESS_WAIT
Sets the -wait N option of cutechess-cli,
that is wait N milliseconds between games,
default=5000 or 5s.
Lakas v0.35.0