J-jHNndZddlZddlmZddlZddlZddlmZddlm Z ddl m Z ddl m Z dd l mZdZ dZ d ZeZ dZ d Zd d ddddddddddddddddddd d!d"d#d$d%d&d'd(d)d*d+d,d-d.d/d0d1d2d3d4d5d6d7d8d9d:d;dd?d@dAdBdCdDdEdFdGdHdIdJdKdLdMdNdOdPdQdRdSe j*dTzdUdVdWdXdYfNdZZeZeed[ej5DZed\eDZdid]Zd^Zd_Z d`Z!daZ"dbDcic] }|e#e|c}Z$GdcddeZ%e%eZGdedfe&Z'Gdgdhe&Z(ycc}w)jz#Parameters and Options for CMA-ES. N)inf)boundary_handler) utilities)utils) CMADataLogger)RecombinationWeightsg? zWTrue # or False or any CMAAdaptSigmaBase class e.g. CMAAdaptSigmaTPA, CMAAdaptSigmaCSAz1 as numbers (rounded), for `True` about 0.16 is usedz92 # 0=unconditional, 1=selective, 2=selective with delayzFalse #v True is untested, also changes respective parameterszhNone #v poorly tested, [0, 0] means const length N**0.5, [-1, 1] allows a variation of +- N/(N+2), etc.z2False #v use squared length for sigma-adaptation z6False #v pc is invariant and ps (default) is unbiasedz81 #v positive multiplier for step-size damping with TPAz|False #v number of iterations of median worsenings threshold at which the sigma change is stalled; the default may become 2zQBoundTransform # or BoundPenalty, unused when ``bounds in (None, [None, None])``z^[None, None] # lower (=bounds[0]) and upper domain boundaries, each a scalar or a list/vectorzR[1e8, 1e12] # when to alleviate the condition in the coordinates and in main axeszETrue # evaluate the final mean, which is a favorite return candidatezTNone # dictionary with index-value pairs like {0:1.1, 2:0.1} that are not optimizedz,-inf #v target function value, minimizationz[] # index list, invokes basic integer handling by setting minstd of integer variables if it was not given and by integer centeringzeis_feasible #v a function that computes feasibility, by default lambda x, f: f not in (None, np.nan)z.inf #v maximum number of function evaluationszE100 + 150 * (N+3)**2 // popsize**0.5 #v maximum number of iterationszAFalse #v sample two new solutions colinear to previous mean shiftzI0 #v minimal std in any arbitrary direction, cave interference with tol*z]0 #v minimal std (scalar or vector) in any coordinate direction, cave interference with tol*zCNone #v maximal std (scalar or vector) in any coordinate directionzN1/3 # maximal std relative to bound_range per coordinate, overruled by maxstdz4False #v one line sample along the evolution path pczl4 + 3 * math.log(N) # population size, AKA lambda, int(popsize) is the number of new solution per iterationzK1 # multiplier for popsize, convenience option to increase default popsizezpnp.random.randn #v randn(lam, N) must return an np.array of shape (lam, N), see also cma.utilities.math.randhssaNone # deprecated, rather use fitness_transformations.ScaleCoordinates instead (or CMA_stds). WAS: Scale for each variable in that effective_sigma0 = sigma0*scaling. Internally the variables are divided by scaling_of_variables and sigma is unchanged, default is `np.ones(N)`ztime # random number seed for `numpy.random`; `None` and `0` equate to `time`, `np.nan` means "do nothing", see also option "randn"zcma_signals.in # read versatile options from this file which contains a single options dict, e.g. ``{"timeout": 0}`` to stop right away, some string-values are recognized, e.g. "np.inf" is valid, use `None` or `""` for no filez[] #v a function or list of functions returning True for termination, called in `stop` with `self` as argument, could be abused for side effectszjinf #v stop if timeout seconds are exceeded, the string "2.5 * 60**2" evaluates to 2 hours and 30 minuteszR1e14 #v stop if the condition of the covariance matrix is above `tolconditioncov`z1e3 #v termination when step-size increases by tolfacupx (diverges). That is, the initial step-size was chosen far too small and better solutions were found far away from the initial solution x0zy1e20 #v sigma/sigma0 > tolupsigma * max(eivenvals(C)**0.5) indicates "creeping behavior" with usually minor improvementsz?1 #v iterations tolerated with flat fitness before terminationzJ1e-11 #v termination criterion: tolerance in function value, quite usefulzD1e-12 #v termination criterion: tolerance in function value historyzw0 #v termination criterion: relative tolerance in function value: Delta f current < tolfunrel * (median0 - median_min)zaint(100 + 100 * N**1.5 / popsize) #v termination if no improvement over tolstagnation iterationsa[1e-9, 20, 0.1] #v termination thresholds for Delta of [mean, iterations, iterations fraction], the latter two are summed; trigger termination if Dmean stays below the threshold over Diter iterations, pass `False` or a negative value to turn off tolxstagnationz71e-11 #v termination criterion: tolerance in x-changeszTNone # deprecated, use a wrapper like those in cma.fitness_transformations instead.zENone # deprecated, use `cma.fitness_transformations.Shifted` insteadz\None #v number of iterations without distribution update, name is subject to future changesz3 #v verbosity e.g. of initial/final message and logging, -1 is very quiet, <=-9 is maximally quiet, <=-10 ignores the signals_filenamezI0 # initial evaluation counter, if append, do not overwrite output fileszC100 #v verbosity: display console output every verb_disp iterationz/inf #v start overwriting after given iterationz- # output path (folder) and filenames prefixzy1 #v verbosity: write data to files every verb_log iteration, writing can be time critical on fast to evaluate functionszyN * (N <= 50) # allow to execute eigendecomposition for logging every verb_log_expensive iteration, 0 or False for neverz<0 #v in fmin2(): plot() is called every verb_plot iterationz"True #v output timings on consolezW{} #? versatile set or dictionary for hacking purposes, value found in self.opts["vv"]cN(ttS)zuse this function to get keyword completion for `CMAOptions`. ``cma.CMAOptions('substr')`` provides even substring search. returns default options as a `dict` (not a `cma.CMAOptions` `dict`). )dictlocals)N AdaptSigma CMA_activeCMA_active_injected CMA_cmeanCMA_const_trace CMA_diagonalCMA_diagonal_decodingCMA_eigenmethod CMA_elitist!CMA_injections_threshold_keep_len CMA_mirrorsCMA_mirrormethodCMA_muCMA_on CMA_samplerCMA_sampler_options CMA_rankmu CMA_rankoneCMA_recombination_weightsCMA_dampsvec_facCMA_dampsvec_fade CMA_teststdsCMA_stds CSA_dampfacCSA_damp_mueff_exponentCSA_disregard_lengthCSA_clip_length_value CSA_squaredCSA_invariant_path TPA_dampfac+stall_sigma_change_on_divergence_iterationsBoundaryHandlerboundsconditioncov_alleviateeval_final_meanfixed_variablesftargetinteger_variables is_feasible maxfevalsmaxitermean_shift_line_samplesmindxminstdmaxstdmaxstd_boundrangepc_line_samplespopsizepopsize_factorrandnscaling_of_variablesseedsignals_filenametermination_callbacktimeouttolconditioncov tolfacupx tolupsigmatolflatfitnesstolfun tolfunhist tolfunrel tolstagnationtolxstagnationtolxtransformation typical_x updatecovwaitverbose verb_append verb_dispverb_disp_overwriteverb_filenameprefixverb_logverb_log_expensive verb_plot verb_timevvsN d/Users/jameslopez/projects/TradingBot25/.venv/lib/python3.12/site-packages/cma/options_parameters.pycma_default_options_r]sb >c#NK|]\}}|jddkDr|yw)z #v rN)find).0kvs r\ rds,%<5P6Aq()v(:&'5Ps#%c#@K|]}|j|gywN)lower)rass r\rdrdsL8K1A8Ksct||r|n|}tjj||rdndz }t t t |z|z tdz| z fS)acan be reassigned/overwritten like a global "parameter setting" This function returns the minimum of three bounds, an absolute bound (default 0.2), mueff / N, and a bound computed from the normal quantile (when the mean is assumed to be in the domain middle) which takes ``ptarget=integer_lower_bound_target_probability(N, N_int)`` (which should be roughly 1/N) as input. This bound becomes < 0.2 only when ptarget < 1/152 (==> < -2.5-sigma). r?)&integer_lower_bound_target_probabilityrmath normal_ppfmininteger_std_lower_bound_limitinteger_std_lower_bound_factor1integer_std_lower_bound_factor2)NmueffN_intbinaryptargetppfs r\integer_std_lower_boundrysb5QANG .. # #GFq$B CC -/%7!;/#5< r^cX||kDrtdj||dd|z|zz S)atarget probability for an integer mutation assuming a centered mean and no boundaries (two-tailed). ``2 / (N + N_int)`` should keep at least 37% of the solutions unaffected. Details: from ``ptarget`` we can compute ``sigma = 1 / PPF(ptarget/2) / 2`` where ``PPF`` is the quantile function of the standard normal distribution and the first ``2`` accounts for two-sided sampling of the tails and the second ``2`` is needed because the distance to the value domain bound is 1/2. z({0}=N_int > N = {1} is not a valid casesrj) ValueErrorformat)rsrus r\rlrls8 qyCJJ5RSTUU A r^c|dur dti}n|sddi}t|tsd|i}|jdt|jdtj |S)zreturn a `dict` with ``'maxrestarts'`` and ``'maxfevals'`` as keys. `restarts` is a parameter to ``cma.fmin*``, see `cma.fmin`. T maxrestartsrr5)"default_restart_number_if_not_zero isinstancer setdefaultnpr)restartss r\amend_restarts_parameterrsb 4!#EF !1% h %!8,  'IJ  RVV, Or^c:|duxrtj| S)zdefault to check feasibility of f-values. Used for rejection sampling in method `ask_and_eval`. :See also: CMAOptions, ``CMAOptions('feas')``. N)ris_nan)xfs r\r4r4s D= 0a00r^cddlm}|j|jddt ddDj ddj d d S) areturn a string safe to `eval` or raise an exception. Selected words and chars are considered safe such that all default string-type option values from `CMAOptions()` pass. This function is implemented for convenience, to keep the default option format backwards compatible, and to be able to pass, for example, `3 * N`. Function or class names other than those from the default values cannot be passed as strings (any more) but only as the function or class themselves. r)purecma#rc3$K|]}||g ywrf)rarbs r\rdzsafe_str..s !-",-.!Q",s)TrueFalseNonersdimr=intzmath.logznp.infrznp.logznp.random.randntimeBoundTransformr4znp.linalg.eighz{}/zN onerz/ /z//)rsafe_strsplitr replace)rhrs r\rrs[   AGGCLO !-", !- - &ggv((/(= >r^) BoundNone BoundPenaltyrceZdZdZdZdZdZedZedZ ddZ ddZ ddZ dfd Z dd Zdd Zd Zed ZddZdZddZddZddZdZdZdZedZdZddZxZS) CMAOptionsaZa dictionary with the available options and their default values for class `CMAEvolutionStrategy`. ``CMAOptions()`` returns a `dict` with all available options and their default values with a comment string. ``CMAOptions('verb')`` returns a subset of recognized options that contain 'verb' in there keyword name or (default) value or description. ``CMAOptions(opts)`` returns the subset of recognized options in ``dict(opts)``. Option values can be "written" in a string and, when passed to `fmin2` or `CMAEvolutionStrategy`, are evaluated using "N" and "popsize" as known values for dimension and population size (sample size, number of new solutions per iteration). All default option values are given as such a string. Details ------- `CMAOptions` entries starting with ``tol`` are termination "tolerances". For `tolstagnation`, the median over the first and the second half of at least `tolstagnation` iterations are compared for both, the per-iteration best and per-iteration median function value. Example ------- :: import cma cma.CMAOptions('tol') is a shortcut for ``cma.CMAOptions().match('tol')`` that returns all options that contain 'tol' in their name or description. To set an option:: import cma opts = cma.CMAOptions() opts.set('tolfun', 1e-12) opts['tolx'] = 1e-11 todo: this class is overly complex and should be re-written, possibly with reduced functionality. :See also: `fmin2` (), `CMAEvolutionStrategy`, `CMAParameters` FTctS)z>return a dictionary with default option values and description)cma_default_optionsrr^r\defaultszCMAOptions.defaults*s #"r^ctS)acreturn list of options that can be changed at any time (not only be initialized). Consider that this list might not be entirely up to date. The string ' #v ' in the default value indicates a versatile option that can be changed any time, however a string will not necessarily be evaluated again. )cma_versatile_optionsrr^r\versatile_optionszCMAOptions.versatile_options2s %$r^cl|j||j||j||S)z6check for ambiguous keys and move attributes into dict) check_valuescheck_attributes)selfoptionss r\checkzCMAOptions.checkAs1 '" g& '" r^c tj}g}g}||}|D]}||}|Dt|dtt t |dtt t ||vr*||k(r||j |}t|d|d|j||j||S)Nz2 is not a valid option. Similar valid options are z Valid options are z was not a unique key for z option)r corrected_keyr{strlistrindexappend)rrrvalidated_keys original_keyskey correct_keys r\rzCMAOptions.check_valuesGs" 22  ?GC',K" "%c$/B3/G*H&I!$T*=%>!? "ABB n,+%'(<(|j ds-tt| tj|d}net|tr!|r tdtt| |n4|r'||j drtt| |n tdt j |r|j ds|jt|jD]{}|j|}|tjvr8t jdt!|zdzd d |j#|b||k7sh|j#|||<}d |_|j&j)|_t-|j*|j*d <y) aIreturn an `CMAOptions` instance. Return default options if ``s is None and not kwargs``, or all options whose name or description contains `s`, if `s` is a (search) string (case is disregarded in the match), or with entries from dictionary `s` as options, or with kwargs as options if ``s is None``, in any of the latter cases not complemented with default options or settings. Returns: see above. Details: as several options start with ``'s'``, ``s=value`` is not valid as an option setting. Nnocheck uncheckedz-Dictionary argument must be the only argumentzMThe first argument must be a string or a dict or a keyword argument or `None`)rrzinvalid key ``z `` removed__init__rFr)superrrrris_str startswithmatchrr r{rrrrrrpop _lock_settingrcopyrlen)rrhkwargsrr __class__s r\rzCMAOptions.__init__s& 9V *d ,Z-@-@-B CA \\!_Q\\+%> *d ,Z\-?-?-B CA 4  !PQQ *d ,Q / all;&? *d ,V 4lm m||Aall3K&L JJLDIIK("005 j&9&9&;;''(83s8(C#/)01;\KHHSMK'(, D%)#==--/*-d.>.>*?'r^c|j||}|||i}|jD]O\}}|j|}|tj vr|s0t dt |zdzK|||<Q|S)ainitialize one or several options. Arguments --------- `dict_or_str` a dictionary if ``val is None``, otherwise a key. If `val` is provided `dict_or_str` must be a valid key. `val` value for key Details ------- Only known keys are accepted. Known keys are in `CMAOptions.defaults()` z&Warning in cma.CMAOptions.init(): key z ignored)ritemsrrrprintr)r dict_or_strvalwarndicrs r\initzCMAOptions.inits" ; ?s#C HC$$S)C*--//BC!#-./ S $ r^c(|||i}tt|jD]f\}}|j|}|jr|t j vs|r|||<Atjdt|zdzddh|S)adassign versatile options. Method `CMAOptions.versatile_options` () gives the versatile options, use `init()` to set the others. Arguments --------- `dic` either a dictionary or a key. In the latter case, `val` must be provided `val` value for `key`, approximate match is sufficient `force` force setting of non-versatile options, use with caution This method will be most probably used with the ``opts`` attribute of a `CMAEvolutionStrategy` instance. zkey z& ignored (not recognized as versatile)setr) rr rrrrrrrr)rrrforce key_originalrs r\rzCMAOptions.sets( ?s)C!%d3ioo&7!8 L#$$\2C&&z3355S ##FS->$>>%?$l4"9 r^c|jtjD]!}||vstj|||<#|S)z1add all missing options with their default values)rrr)rrs r\ complementzCMAOptions.complementsE &&(C$&//1#6S ) r^c fttdt|jDS)zreturn the subset of those options that are settable at any time. Settable options are in `versatile_options` (), but the list might be incomplete. c3RK|]}|dtjvr|!ywrN)rr)rais r\rdz&CMAOptions.settable..s.K*C !48G,j#>  5 #::c? " #&  q6$$[1s1v&   sSC(># J iik * *+C0 03K%%c*}%)$$[1 4 $'}#5#:}QDDr^cL|r|j|}||||||<||S)abEvaluates and sets the specified option value in environment `loc`. Many options need ``N`` to be defined in `loc`, some need `popsize`. Details ------- Keys that contain 'filename' are not evaluated. For `loc` is None, the self-dict is used as environment :See: `evalall()`, `__call__` )r)rrrrrs r\rzCMAOptions.evalXs3 $$S)Cgs+S Cyr^c(|j| t}d|vrj|d|d|}t|jD]A}|j |}|j dr&|j ||||d|dCd|_|S)zSEvaluates all option values in environment `loc`. :See: `eval()` rsr=rrsr=T)rr]rrrrrr)rrrr=rbs r\evalallzCMAOptions.evalallls  +-H #:9hy&93?G$))+&&&q)<<$ !Xa["3x7;= ' " r^c|j}i}t|D]=}t|dzt||zdz}||jvs6||||<?t|S)zreturn all options that match, in the name or the description, with string `s`, case is disregarded. Example: ``cma.CMAOptions().match('verb')`` returns the verbosity options. ='')rgsortedrr)rrhrresrbs r\rzCMAOptions.matchsh AAT!W-4A !aA#r^c 8 t|d|d<|dr|j||dsyt|d|kDr't dj |t|d|dr&t|ddk(r |dddg|d<d|dd<|j dddtdfvrCd d tj|tjt|dd zzzz|d<|j d ddtd fvrd d t|d|z zz|d <t|d|k(rF|j dddtdfvrd|d<|j dddtdfvrd|d<yyy#t$r|drg|d<YwxYw)z;amend options when integer variables are indicated r3NzP{0} = dimension < len(options['integer_variables']) = {1} is not a valid settingr/rrr=rrHrIrJ) rramend_integer_variablesrr{r|getrrmlog)r dimensioninoptss r\amend_integer_optionsz CMAOptions.amend_integer_optionss +(,T2E-F(GD$ % # $  ( ( 3'(  t'( )I 57$fYD9L4M0NOQ Q ( )4012a7267O2PQR2SUV1W-.12D) *2 . ::i &41DY1O*P P!txx ':'+xxD9L4M0NQT0T'U(V#WWDO ::& --.>?2A A%&D,-.:*<&t j d||fz|dj|dj|yy)zset lower std bounds for integer variables. Uses the above defined `integer_std_lower_bound` function which can be reassigned. r3Nr9rz?dropping integer index %d as it is not in range of dimension %d) risscalaronesryrrrrr)rrsrtrs r\set_integer_min_stdzCMAOptions.set_integer_min_stds '(  ;;tH~ &4>Q+>!(^bggaj8DN{{4>*-.2;Q;H~a(A-,C !5#d3F.G*H-JXq)''Y !f%&,-11$7J2K2Q2QRS2TU/+r^ct)z_return options as const attributes of the returned object, only useful for inspection. )NotImplementedErrorrs r\ to_namedtuplezCMAOptions.to_namedtuples "!r^c@|j|jS)z[update options from a `collections.namedtuple`. :See also: `to_namedtuple` )r_asdict)rts r\from_namedtuplezCMAOptions.from_namedtuples{{199;''r^c~t|jD]}t|ddzt|dzdz}|jd}d}|s<|rTt |t |dz|kr7|d|j dzz }|rt |t |dz|kr7t |d}|rfy)Nrrrr rz )rrrrrrr)r linebreakrrhalines r\pprintzCMAOptions.pprint s %AAaD D 3qt9,s2A ADCIAaD 1I=C!%%(N*DCIAaD 1I=d !  &r^rf)NTNF)NN)NNT)r)P)__name__ __module__ __qualname____doc__ _ps_for_pc_hsig_stationary_sphere staticmethodrrrrrrrrrpropertyrrrrrrrrrrrr  __classcell__)rs@r\rrs2fJ E## % % ,+Z1@f D D L L-^E$(* 0'd)@V4""( "r^rc&eZdZdZddZddZdZy) CMAParametersaMstrategy parameters like population size and learning rates. Note: contrary to `CMAOptions`, `CMAParameters` is not (yet) part of the "user-interface" and subject to future changes (it might become a `collections.namedtuple`) Example ------- >>> import cma >>> es = cma.CMAEvolutionStrategy(20 * [0.1], 1) #doctest: +ELLIPSIS (6_w,12)-aCMA-ES (mu_w=3.7,w_1=40%) in dimension 20 (seed=...) >>> >>> type(es.sp) # sp contains the strategy parameters >>> es.sp.disp() #doctest: +ELLIPSIS {'CMA_on': True, 'N': 20, 'c1': 0.00437235..., 'c1_sep': ...0.0343279..., 'cc': 0.171767..., 'cc_sep': 0.252594..., 'cmean': 1.0, 'cmu': 0.00921656..., 'cmu_sep': ...0.0565385..., 'lam_mirr': 0, 'mu': 6, 'popsize': 12, 'weights': [0.4024029428..., 0.2533890840..., 0.1662215645..., 0.1043752252..., 0.05640347757..., 0.01720770576..., -0.05018713636..., -0.1406167894..., -0.2203813963..., -0.2917332686..., -0.3562788884..., -0.4152044225...]} >>> :See: `CMAOptions`, `CMAEvolutionStrategy` c^||_|dk(r|d}d|_ |j|||y)zCompute strategy parameters, mainly depending on dimension and population size, by calling `set`. When ``ccovfac == 1``, we use ``ccovfac=opts['CMA_on']``. rrN)ccovfacrR)rsr=r)rrsrr0rRs r\rzCMAParameters.__init__Hs7  a<8nG P w8r^Nc d}d}d}|}|j} |r|j| |dn|jd| id}||dz}t|| dzzd z }tj|d r)t |d |_t|j }n|d rxt d |d z|_t|j |krZ|j j|j jdt|j |krInt ||_||_ |j j|_ |d dkrtd|d |zz|_ nI|d d kDrtd|d z|_ n*tddt|d | zd zfzzdz|_ |jd z|jd |jzz d zkDr]tjdd|jd z|jd |jzz d z|j|jfzz|j|jd zkDr td|j j} d} d|| | z z| zz| | z|d | z| z z| zzz |_d d | z z| | z z| dzd | z zd | z| z zz |_t%|ddrld|dvr(dd| | z zdzz| dzdzd | z| z dzzz |_d|dvr6|dd|_|j |_t'd|j zd|dz|ztd |jdz zd z| dzdz| zz |_t+|d|z|| | | z|_dd krd|_t'd!tj|d"rd} d#} t%|ddrd$|dvr|dd$} t'd%| z| }td |j(z |d"|z| z| |zd |z zd z z| d zdz| |zd z zz |_t%|ddrd&|dvr |dd&|_t+td |j,z ||| | | z|_nd'x|_|_t%|ddrd|dvr |dd|_t3d(|j Dr[|d)r<|d*r7|d"r2|j j5| |j(|j.n|j j7|j(|j.zd'kD|_|d*s|d*d+gd,d-fvrd.|_dd krAd/d t;| |jz | d z | d zz dzd z gzz|j<z|_dd kr`|d0d d'|jz|jz z z| z|jz d/z|j<z|_t'd1|j>dd kr,d2|j>z|_t'd3|j>zt+|d4|_ dd krCtCj2|j@d k7r!t'd5tE|j@z|r@|j8s t'd6|d #t'd7|j jzy+y+y+)8zSCompute strategy parameters as a function of dimension and population size g@cNdt|d|dzzzt||z zz S)*used for computing separable learning rate?@rkfloat)dfmurss r\conedfz!CMAParameters.set..conedf[s,b2C</%)a-?@@ @r^cTt||zd|z zdz |d|dzzz|dz zz S)r3r4rjrkr5r6)r8r9alphamus r\cmudfz CMAParameters.set..cmudf_s;'B,b014a"c'k9IBQSG9STU Ur^rrsr=r>rrr rrjrrkg{Gz?g(\?z1pairwise selection is not implemented, therefore zG mu = %d > %d = %d - 2*%d + 1 = popsize - 2*mirr + 1 can produce a biaszzfraction of mirrors in the population as read from option CMA_mirrors cannot be larger 0.5, theoretically optimal is 0.159r4r[ __getitem__ sweep_ccov1r<sweep_cczcc is %frrg?r5rrz c1 is zerorg?sweep_rankmu_offsetzrankmu_offset = %.2f sweep_ccovrc3&K|] }|dk ywrr)raws r\rdz$CMAParameters.set..s)jq1ujsrrNrrFg333333?r%zdamps = z damps is %frz cmean = z'covariance matrix adaptation turned offzmu = %d)#rsrrris_r weightsrinsertr9r=lam_mirrrorr{rtcccc_seprrc1r7c1_sepcmucmu_sepanyfinalize_negative_weightszero_negative_weightsrmaxcsdampscmeanrr)rrr=r0rR limit_fac_ccr:r>sprsrtbalphacov rankmu_offsetr9s r\rzCMAParameters.setUs#  A V DD  LLa73 4llC7+I6G t,- -GgC'!+, 99T56 7-d3N.OPBJ"**oG (^-a$x..@ABJbjj/G+ !!"**--5bjj/G+.g6BJ     $cD$7'$AABBK - 1 $cD$778BKcD3Q/C+D$DDtKLBK ::?RZZ!bkk/9A= =    S[JJ!ORZZ!bkk/%AA%ErzzSUS^S^_``!` a ;;q (:; ;    |eai/!33AE A 599;QY*Va!e^a%i!m35 4:} -T *q519}s22q1usl56Y]S4H7IJT$Z'T :.EE j255()tM**W4s1bjj1n7MM3w#o-/$}-7&E1:MMN 6BE ,  99T,' (H MtDz=16KtTXz6Y $T +@ A ,}<=BRUUl+g5@'",q2v59;q5C-(R-!*;; =>BFtDz=1ld4j6PdL1s1ryy='E!UMrsd !7"#K"#7 #Q6:3Q%&"iM\5c;X]b&u?P I >X_XZu{U~ ~haYEDOJ1_g kp[j :]w> S_ U j PfJ z` |o H8b xhA-T WUKuQ CiVp`[SI%447ffA?L2`Sqf+,f%<5H5N5N5P%<<=L8KLL"" 1>8   /3D'"D ))   i"i"V!!5!78^$F^$BZ,VZ,g s/D2