&`ii ddlZddlZddlmZmZmZmZmZmZm Z ddl Z ddl Z ddl mZmZmZddlmZmZddlmZmZddlmZmZmZmZmZmZmZmZm Z m!Z!ddl"m#Z#m$Z$e\Z%Z&Z%e\Z'Z%ej(dZ)eeeee!eeeed Z*e#dd ee+efd e,d e fdZ-e$dGddZ.GddZ/dS)N)AnyDictListOptionalTupleTypeUnion)DEPRECATED_VALUE Deprecateddeprecation_warning) deep_update force_tuple) try_import_tftry_import_torch) EmaStatsItemSeriesStats ItemStatsLifetimeSumStatsMaxStats MeanStatsMinStatsPercentilesStats StatsBaseSumStats) DeveloperAPI PublicAPIz ray.rllib) meanemaminmaxsum lifetime_sum percentilesitem item_seriesFstateis_rootreturnc|d}| |gd}|dddur|dkrod|d<|rtj|d |d<nd |d<|d d|d <nd|d <td}||}|St dd|dkrQ|d0d}|d|d<tj|d |d<d|d<nd}|d |d <|d|d<n|dvr=|d |d <|d|d<|dkr|d d|d<npd|d<nj|A|dddur)d}|d |d <|d|d<|d|d<d|d<n'|dkr!|d |d <|d|d<|d|d<t|}||d<||}|S)z+Creates a Stats object from a legacy state.reduce)is_leafr' latest_mergedclear_on_reduceTFr!r"stats_cls_identifiervaluesgthroughput_statsNtrack_throughputsr&zOLegacy Stats class tracking throughput detected. This is not supported anymore.errorr ema_coeffrvaluewindow)rr r!track_throughputr#)getnpnansumDEFAULT_STATS_CLS_LOOKUP from_stater nanmean)r&r'cls_identifier new_state_clsstatss z/home/jaya/work/projects/VOICE-AGENT/VIET/agent-env/lib/python3.11/site-packages/ray/rllib/utils/metrics/metrics_logger.pystats_from_legacy_staterD.s8_N; I  yy"D))U22 U " "0>I, - 0,.IeHo,F,F .)),/ .)yy+,,815 -..16 -.+N;DOO)O44EL a        )"N%*;%7Ik "!#E(O!d,d?Z*ed=d>d,d@Z+d S)F MetricsLoggerajA generic class collecting and reducing metrics. Use this API to log and merge metrics. Metrics should be logged in parallel components with MetricsLogger.log_value(). RLlib will then aggregate metrics, reduce them and report them. The MetricsLogger supports logging anything that has a corresponding reduction method. These are defined natively in the Stats classes, which are used to log the metrics. Please take a look ray.rllib.utils.metrics.metrics_logger.DEFAULT_STATS_CLS_LOOKUP for the available reduction methods. You can provide your own reduce methods by extending ray.rllib.utils.metrics.metrics_logger.DEFAULT_STATS_CLS_LOOKUP and passing it to AlgorithmConfig.logging(). Notes on architecture: In our docstrings we make heavy use of the phrase 'parallel components'. This pertains to the architecture of the logging system, where we have one 'root' MetricsLogger that is used to aggregate all metrics of n parallel ('non-root') MetricsLoggers that are used to log metrics for each parallel component. A parallel component is typically a single Learner, an EnvRunner, or a ConnectorV2 or any other component of which more than one instance is running in parallel. We also allow intermediate MetricsLoggers that are no root MetricsLogger but are used to aggregate metrics. They are therefore neither root nor leaf. Fstats_cls_lookupci|_t|_||_t j|_||_dS)aWInitializes a MetricsLogger instance. Args: root: Whether this logger is a root logger. If True, lifetime sums (reduce="lifetime_sum") will not be cleared on reduce(). stats_cls_lookup: A dictionary mapping reduction method names to Stats classes. If not provided, the default lookup (ray.rllib.utils.metrics.metrics_logger.DEFAULT_STATS_CLS_LOOKUP) will be used. You can provide your own reduce methods by extending ray.rllib.utils.metrics.metrics_logger.DEFAULT_STATS_CLS_LOOKUP and passing it to AlgorithmConfig.logging(). N)rB _DummyRLock_threading_lock_is_root_loggertime perf_counter_time_when_initializedrJ)selfrootrJs rC__init__zMetricsLogger.__init__s@  +}}#&*&7&9&9# 0rEkey.r(c,||S)a&Returns True, if `key` can be found in self.stats. Args: key: The key to find in self.stats. This must be either a str (single, top-level key) or a tuple of str (nested key). Returns: Whether `key` could be found in self.stats. ) _key_in_stats)rRrUs rC __contains__zMetricsLogger.__contains__s!!#&&&rENTcompile throughputlatest_merged_onlycJ|r(|js Jd|||Sdtttfffd }|j5|||jcdddS|||d}n||d}t|tr#|  cdddSt|tr|r||cdddS|cdddS#1swxYwYdS) aReturns the reduced values found in this MetricsLogger. Note that calling this method does NOT cause an actual underlying value list reduction, even though reduced values are being returned. It'll keep all internal structures as-is. By default, this returns a single reduced value or, if the Stats object has no reduce method, a list of values. When when compile is False, the result is a list of one or more values. Args: key: The key/key sequence of the sub-structure of `self`, whose (reduced) values to return. default: An optional default value in case `key` cannot be found in `self`. If default is not provided and `key` cannot be found, throws a KeyError. compile: If True, the result is compiled into a single value if possible. throughput: If True, the throughput is returned instead of the actual (reduced) value. latest_merged_only: If True, only considers the latest merged values. This parameter only works on aggregation loggers (root or intermediate). Returns: The (reduced) values of the (possibly nested) sub-structure found under the given key or key sequence. z0Throughput can only be peeked from a root logger)rUdefaultrBc|dtdtffd }tj||S)NpathrBc  |S#t$r}td|dd|d|d}~wwxYw)NrYr[zError peeking stats  with compile= at path .)peek Exception ValueError)r_rBerYr[s rC_peek_with_pathzAMetricsLogger.peek.._nested_peek.._peek_with_paths~ :: 'A)strrtreemap_structure_with_pathcopy)rBrirYr[s rC _nested_peekz(MetricsLogger.peek.._nested_peeksQ c )       /NN NrENT key_errorFra) rN_get_throughputsrrjrrMrB_get_key isinstancerredict)rRrUr]rYrZr[rnrBs ` ` rCrezMetricsLogger.peeks>  C$ B BA B B$((S'(BB B OS#X O O O O O O O ! # #{#|DJ// # # # # # # # #? MM#M>>EE MM#M??EeY// # :: 'z,MetricsLogger.peek_results..s)Ay1I1IPaffWf---qrE)rk map_structure)rurYs `rC peek_resultszMetricsLogger.peek_resultss)! P P P P    rErr*r7r5r#r-with_throughputthroughput_ema_coeffreduce_per_index_on_aggregater*r7r5r#r-r~rrkwargsc |j5||s|dkr|d}|r|dkstd|dkr|tddd}| turtd d |turtd d|d kr|td d|r|dvrtdd  |j|} n!#t $rtd|dwxYw||| d<||| d<||| d<||| d<|j| d<d| d<| di| } ||| ddddS#1swxYwYdS)zCPrepare the kwargs and create the stats object if it doesn't exist.rNg{Gz?r#z4percentiles is only supported for reduce=percentileszcwindow is not supported for ema reduction. If you want to use a window, use mean reduction instead.Tr3zreduce_per_index_on_aggregate is deprecated. Aggregation now happens over all valuesof incoming stats objects, treating each incoming value with equal weight.Fzthroughput_ema_coeff is deprecated. Throughput is not smoothed with ema anymorebut calculate once per MetricsLogger.reduce() call.rzJema_coeff is not supported for mean reduction. Use `reduce='ema'` instead.)r!r"zwith_throughput=True is only supported for reduce='sum' or reduce='lifetime_sum'. Use reduce='sum' or reduce='lifetime_sum' instead.zInvalid reduce method 'z(' could not be found in stats_cls_lookupr7r5r~r'r+) rMrWrgr r rJKeyErrorrN_set_key) rRrUr*r7r5r#r-r~rrr stats_cls stats_objects rC_maybe_create_stats_objectz(MetricsLogger._maybe_create_stats_object s !C 1C 1%%c**A 1U??y'8 $Iv'>'>$NU??v'9'}""F08HHH'e# (/???'N" V## ,+h"& #v5L'L'L'_# $ 5f =II$b&bbb %'-F8$(*3F;'*,7F=)".0?F,-%)$8y!%)y!(y22622  c<000GC 1C 1C 1C 1C 1C 1C 1C 1C 1C 1C 1C 1C 1C 1C 1C 1C 1C 1s+B7E C EC,,A EE  E r6c ||d}|d}|turtdd|dkr|durd}d}||||||||| | ||} | |dS) ax Logs a new value or item under a (possibly nested) key to the logger. Args: key: The key (or nested key-tuple) to log the `value` under. value: A numeric value, an item to log or a StatsObject containing multiple values to log. reduce: The reduction method to apply when compiling metrics at the root logger. By default, the reduction methods to choose from here are the keys of rllib.utils.metrics.metrics_logger.DEFAULT_STATS_CLS_LOOKUP. You can provide your own reduce methods by extending rllib.utils.metrics.metrics_logger.DEFAULT_STATS_CLS_LOOKUP and passing it to AlgorithmConfig.logging()). window: An optional window size to reduce over. If not None, then the reduction operation is only applied to the most recent `window` items, and - after reduction - the internal values list under `key` is shortened to hold at most `window` items (the most recent ones). Must be None if `ema_coeff` is provided. If None (and `ema_coeff` is None), reduction must not be "mean". ema_coeff: An optional EMA coefficient to use if `reduce` is "mean" and no `window` is provided. Note that if both `window` and `ema_coeff` are provided, an error is thrown. Also, if `ema_coeff` is provided, `reduce` must be "mean". The reduction formula for EMA is: EMA(t1) = (1.0 - ema_coeff) * EMA(t0) + ema_coeff * new_value Defaults to 0.01. percentiles: If reduce is `None`, we can compute the percentiles of the values list given by `percentiles`. Defaults to [0, 0.5, 0.75, 0.9, 0.95, 0.99, 1] if set to True. When using percentiles, a window must be provided. This window should be chosen carefully. RLlib computes exact percentiles and the computational complexity is O(m*n*log(n/m)) where n is the window size and m is the number of parallel metrics loggers involved (for example, m EnvRunners). clear_on_reduce: Deprecated. Use reduce="lifetime_sum" instead. If True, all values under `key` will be cleared after `self.reduce()` is called. Setting this to True is useful for cases, in which the internal values list would otherwise grow indefinitely, for example if reduce is None and there is no `window` provided. with_throughput: Whether to track a throughput estimate together with this metric. This is supported by default only for `reduce=sum` and `reduce=lifetime_sum`. throughput_ema_coeff: Deprecated argument. Throughput is not smoothed with ema anymore but calculate once per MetricsLogger.reduce() call. reduce_per_index_on_aggregate: Deprecated argument. Aggregation now happens over all values of incoming stats objects once per MetricsLogger.reduce() call, treating each incoming value with equal weight. Nrrzrclear_on_reduce is deprecated. Use reduce='lifetime_sum' for sums. Provide a custom reduce method for other cases.Fr3r!r"r})r r rrrpush) rRrUr6r*r7r5r#r-r~rrrrBs rC log_valuezMetricsLogger.log_value]sv >f0F >F "2 2 2 E     U??%77#F"O '' #++!5*G (  c"" 5rEr) rUr*r7r5r#r-r~rrc  t|tsJd|dt|   f d} j5t j| |ddddS#1swxYwYdS)aLogs all leafs of a possibly nested dict of values or Stats objects to this logger. Traverses through all leafs of `stats_dict` and - if a path cannot be found in this logger yet, will add the `Stats` found at the leaf under that new key. If a path already exists, will merge the found leaf (`Stats`) with the ones already logged before. This way, `stats_dict` does NOT have to have the same structure as what has already been logged to `self`, but can be used to log values under new keys or nested key paths. Passing a dict of stats objects allows you to merge dictionaries of stats objects that have been reduced by other, parallel components. See MetricsLogger.log_value for more details on the arguments. z`stats_dict` (z) must be dict!c  ttj|z} ||   dS)N) r6r*r7r5r#r-r~rr)rrkflattenr) r_ stat_or_value extended_keyr-r5r# prefix_keyr*rrRrr7r~s rC_mapz$MetricsLogger.log_dict.._mapsb% DL4F4F(G(GGL NN##' / /%9.K      rEN)rsrtrrMrkrl) rR value_dictrUr*r7r5r#r-r~rrrrs ` ```````` @rClog_dictzMetricsLogger.log_dicts8    8 8 7J 7 7 7 8 8 !%%                 ! ; ;  (z : : : ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ;sA55A9<A9 aggregate)newr4c|j|i|SN)rrRargsrs rCmerge_and_log_n_dictsz#MetricsLogger.merge_and_log_n_dictsst~t.v...rE)rU stats_dictscT td fd fd g}|D]0}  |} |}||1D]Њfd|D}jd}t |t rdn|}|F|djdd  }|jr| j n|j rJd d | |  |dS)aMerges n stats_dicts and logs result by merging on the time axis with existing stats. The n stats_dicts should be generated by n parallel components such that merging their respective stats in parallel is meaningful. Stats can be aggregated at root or intermediate loggers. This will replace most internal values with the result of the merge. For exceptions, see the documentation of the individual stats classes `merge` methods. Args: stats_dicts: List of n stats dicts to be merged and then logged. key: Optional top-level key under which to log all keys/key sequences found in the n `stats_dicts`. rcLt|tr1i}|D]\}}|||fz||<|St|tr2|t |dkr|dS|S||SNr)rsrtitemslistaddlen)dr_new_dictrUr6all_keystraverse_and_add_pathss rCrz7MetricsLogger.aggregate..traverse_and_add_pathss!T"" "#''))QQJC$:$:5$#-$P$PHSMMAt$$  T"""q66Q;;Q4K T"""rEct|tr||iSt|dkr|d||ddiS|d|iSr)rsrjr) stats_dictrUbuild_nested_dicts rCrz2MetricsLogger.aggregate..build_nested_dictsc#s## ,Z((SAA 1 1*c!""g F FGGA ++rENclg|]0}||1S)rB)rWrr).0ryrUrRs rC z+MetricsLogger.aggregate..&sR%%c%33 c ++rEFrBrpr)r'r+)init_overrideszCannot aggregate into key 'zg' because it was created by direct logging. Aggregation keys must be separate from direct logging keys.)incoming_stats)r)setappendrrrBrsrtclonerNhas_throughputs$initialize_throughput_reference_timerQr+merger) rRrrUincoming_stats_dicts_with_keyrrstructure_under_key own_statsrrrs ` ` @@@rCrzMetricsLogger.aggregates$55         , , , , ,)+%% = =J..z3?? // ;;J ) 0 0 < < < <! *! *C6N #'--4:QV-"W"W ##6==VCV  +1-33/3/CPU#V#V4 ,BB3")^^^^^^^^) OO>O : : : MM#y ) ) ) )C! *! *rEc h||||||||||  ||S)a Measures and logs a time delta value under `key` when used with a with-block. Args: key: The key (or tuple of keys) to log the measured time delta under. reduce: The reduction method to apply when compiling metrics at the root logger. By default, the reduction methods to choose from here are the keys of rllib.utils.metrics.metrics_logger.DEFAULT_STATS_CLS_LOOKUP. You can provide your own reduce methods by extending rllib.utils.metrics.metrics_logger.DEFAULT_STATS_CLS_LOOKUP and passing it to AlgorithmConfig.logging()). window: An optional window size to reduce over. If not None, then the reduction operation is only applied to the most recent `window` items, and - after reduction - the internal values list under `key` is shortened to hold at most `window` items (the most recent ones). Must be None if `ema_coeff` is provided. If None (and `ema_coeff` is None), reduction must not be "mean". ema_coeff: An optional EMA coefficient to use if `reduce` is "mean" and no `window` is provided. Note that if both `window` and `ema_coeff` are provided, an error is thrown. Also, if `ema_coeff` is provided, `reduce` must be "mean". The reduction formula for EMA is: EMA(t1) = (1.0 - ema_coeff) * EMA(t0) + ema_coeff * new_value percentiles: If reduce is `None`, we can compute the percentiles of the values list given by `percentiles`. Defaults to [0, 0.5, 0.75, 0.9, 0.95, 0.99, 1] if set to True. When using percentiles, a window must be provided. This window should be chosen carefully. RLlib computes exact percentiles and the computational complexity is O(m*n*log(n/m)) where n is the window size and m is the number of parallel metrics loggers involved (for example, m EnvRunners). clear_on_reduce: Deprecated. Use reduce="lifetime_sum" instead. If True, all values under `key` will be cleared after `MetricsLogger.reduce()` is called. Setting this to True is useful for cases, in which the internal values list would otherwise grow indefinitely, for example if reduce is None and there is no `window` provided. with_throughput: Whether to track a throughput estimate together with this metric. This is only supported for `reduce=sum` and `reduce=lifetime_sum`. The current throughput estimate of a key can be obtained through: `MetricsLogger.peek(key, throughput=True)`. throughput_ema_coeff: Deprecated argument. Throughput is not smoothed with ema anymore but calculate once per MetricsLogger.reduce() call. reduce_per_index_on_aggregate: Deprecated argument. Aggregation now happens over all values of incoming stats objects once per MetricsLogger.reduce() call, treating each incoming value with equal weight. r})rrr) rRrUr*r7r5r#r-r~rrs rClog_timezMetricsLogger.log_timeGsQp '' #++!5*G ( }}S!!!rEcdtdtffd }|j5tj||jcdddS#1swxYwYdS)aFReduces all logged values based on their settings and returns a result dict. Note to user: Do not call this method directly! This should be called only by RLlib when aggregating stats. Args: compile: If True, the result is compiled into a single value if possible. If it is not possible, the result is a list of values. If False, the result is a list of one or more values. Returns: A dict containing all ever logged nested keys to this MetricsLogger with the leafs being the reduced stats. r_rBc  |S#t$r}td|dd|d|d}~wwxYw)NrxzError reducing stats rbrcrd)r*rfrg)r_rBrhrYs rC_reducez%MetricsLogger.reduce.._reduceso ||G|444    ZEZZZZSWZZZ s A=AN)rjrrMrkrlrB)rRrYrs ` rCr*zMetricsLogger.reduces # i       ! E E/DD E E E E E E E E E E E E E E E E E EsAA ArzcUse log_value with reduce='item' or another reduction method with a window of the appropriate size.)rhelpr4cdSrrrs rC set_valuezMetricsLogger.set_values rEcT|j5i|_ddddS#1swxYwYdS)z-Resets all data stored in this MetricsLogger.N)rMrBrRs rCresetzMetricsLogger.resetsv  !  DJ                  s !!rorpc2|||dS)aDeletes the given `key` from this metrics logger's stats. Args: key: The key or key sequence (for nested location within self.stats), to delete from this MetricsLogger's stats. key_error: Whether to throw a KeyError if `key` cannot be found in `self`. Raises: KeyError: If `key` cannot be found in `self` AND `key_error` is True. N)_del_key)rRrprUs rCdeletezMetricsLogger.deletes c9%%%%%rEcifd}|j5tj||jdddn #1swxYwYdiS)zReturns the current state of `self` as a dict. Note that the state is merely the combination of all states of the individual `Stats` objects stored under `self.stats`. c\|d|<dS)N--) get_statejoin)r_rBrs rCrz%MetricsLogger.get_state.._maps'*///*;*;Jtyy ' ' 'rENrB)rMrkrlrB)rRrrs @rCrzMetricsLogger.get_states   < < < < < ! ; ;  (tz : : : ; ; ; ; ; ; ; ; ; ; ; ; ; ; ;$$s 7;;r&c|j5i|_|dD]\}}d|vrC|d}||jvsJd|d|j|}||}nt ||j}||d| ddddS#1swxYwYdS) zuSets the state of `self` to the given `state`. Args: state: The state to set `self` to. rBr.zStats class identifier z not found in stats_cls_lookup. This can happen if you are loading a stats from a checkpoint that was created with a different stats class lookup.r2)r&r'rN) rMrBrrJr=rDrNrsplit)rRr&flat_key stats_stater?rArBs rC set_statezMetricsLogger.set_statesS  ! ; ;DJ).w)=)=)?)? ; ;%+)[88%01G%HN&$*????TTTT@??0@D OO+O>>EE4)43GE hnnT22E::::! ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ;sB.CC Crcttj|}||n|j}|D]}||vrdS||}dS)NFT)rrkrrB)rRrrB_dictrUs rCrWzMetricsLogger._key_in_statssZt|H5566*   C%uu#JEEtrErcttj|}||n|j}|D]*} ||} #t$r}|r|icYd}~cSd}~wwxYw|Sr)rrkrrBr)rRrrBrprrUrhs rCrrzMetricsLogger._get_keyst|H5566*   C c    GIIIIIIII    s; AA AAc6ttj|}|j5|j}t |D]?\}}|t |dz kr|||<ddddS||vri||<||}@ ddddS#1swxYwYdSNr)rrkrrMrB enumerater)rRrrBrirUs rCrzMetricsLogger._set_keyst|H5566  ! # #JE#H-- # #3H )))!&E#J # # # # # # # #e##!#E#Jc  # # # # # # # # # # # # # # # # # # #s8B.BBBcXttj|}|j5|j} t |D]4\}}|t |dz kr||=ddddS||}5n#t$r}|r|Yd}~nd}~wwxYwddddS#1swxYwYdSr)rrkrrMrBrrr)rRrrprrrUrhs rCrzMetricsLogger._del_keysNt|H5566  !  JE '11''FAsCMMA---!#J        "#JEE '    G                   s@B.A7- A76B7 BB B BBB#&B#cd}|j5|||d}t|tr-|jst d|d|jcdddS|ikr|cdddS||cdddSifd}tj||j rn|cdddS#1swxYwYdS)aReturns throughput values for Stats that have throughput tracking enabled. If no key is provided, returns a nested dict containing throughput values for all Stats that have throughput tracking enabled. If a key is provided, returns the throughput value for that specific key or nested structure. The throughput values represent the rate of change of the corresponding metrics per second. For example, if a metric represents the number of steps taken, its throughput value would represent steps per second. Args: key: Optional key or nested key path to get throughput for. If provided, returns just the throughput value for that key or nested structure. If None, returns a nested dict with throughputs for all metrics. default: Default value to return if no throughput values are found. Returns: If key is None: A nested dict with the same structure as self.stats but with "_throughput" appended to leaf keys and throughput values as leaf values. Only includes entries for Stats objects that have throughput tracking enabled. If key is provided: The throughput value for that specific key or nested structure. cd}i}tj|D]T\}}|||\}}t|tr+|}|ddD]}||vri||<||}|||d<U|S)z@Helper function to calculate throughputs for a nested structure.ct|tr4|jr-t|}|dd|ddzfz|jfS||fS)N _throughputsrsrrr throughputs)r_r6rUs rC _transformzOMetricsLogger._get_throughputs.._nested_throughputs.._transform@s^eY//UE4IU%d++Css8s2w'?&AA5CTTTU{"rENr)rkflatten_with_pathrsfloat) rBrresultr_r6new_path new_valuerks rC_nested_throughputsz;MetricsLogger._get_throughputs.._nested_throughputs=s # # #F#5e<< 4 4 e&0ju&=&=#)i//4"E%crc]))E>>')E!H %a*3E(2,'MrENFrozKey 'z+' does not have throughput tracking enabledct|tr_|jrZt|}|dd|ddzfz}}|ddD]}||vri||<||}|j||d<dSdSdS)Nr _throughputr)r_rBrUrrrs rCrz,MetricsLogger._get_throughputs.._maphseY// 7E4I 7%d++Ccrc(c"g &=%??C'E "X))E>>')E!H %a%*%6E#b'NNN 7 7 7 7rE) rMrrrsrrrgrrkrlrB)rRrUr]rrBrrs @rCrqzMetricsLogger._get_throughputs#s4   . !$ ;$ ; cU ;;eY// 6 0(TCTTT!,$ ;$ ;$ ;$ ;$ ;$ ;$ ;$ ;b[["$ ;$ ;$ ;$ ;$ ;$ ;$ ;$ ; /.u55!$ ;$ ;$ ;$ ;$ ;$ ;$ ;$ ;$K 7 7 7 7 7  (tz : : :"-:;;7I$ ;$ ;$ ;$ ;$ ;$ ;$ ;$ ;$ ;$ ;$ ;$ ;$ ;$ ;$ ;$ ;$ ;$ ;s$AC(C< C&CC  C c|}|d}t||pidfd|S)aHCompiles all current values and throughputs into a single dictionary. This method combines the results of all stats and throughputs into a single dictionary, with throughput values having a "_throughput" suffix. This is useful for getting a complete snapshot of all metrics and their throughputs in one call. Returns: A nested dictionary containing both the current values and throughputs for all metrics. The structure matches self.stats, with throughput values having "_throughput" suffix in their keys. Trx)new_keys_allowedct|tr,i}|D]\}}|||<|St|trt |dkr|dS|S|Sr)rsrtrrr)rrrUr6 traverse_dicts rCrz,MetricsLogger.compile..traverse_dicts!T"" "#''))99JC$1M%$8$8HSMMAt$$ q66Q;;Q4KrE)rqr*r )rRrr/rs @rCrYzMetricsLogger.compilezsp++-- T**FK-2EEEE     }V$$$rEz$Tensor mode is not required anymore.cdSrrrs rCactivate_tensor_modez"MetricsLogger.activate_tensor_mode rEcdSrrrs rCdeactivate_tensor_modez$MetricsLogger.deactivate_tensor_moderrE)NNTFF)TF)r(N)NN),__name__ __module__ __qualname____doc__r<rrrjrrrTr rboolrXrre staticmethodr|r intrrrrrr rrrr*rrrrrrWrrrrrqrYrrrrErCrIrIts)* % 11# d9o% & 11118 'c5c?&: ; ' ' ' ' '26 #( E#E# 3c3h- .E# E#  E# ! E# E#E#E#E#N  c D C   \ (.2%)8<*:*.0@8HR1R1R1 3c3h' (R1 R1 sEz*+ R1 E? R1eDItO45R1"$R1"$R1'uoR1(0~R1sCx.R1 R1R1R1R1r!%.2%)8<*:*.0@8HWWW 3c3h' (WW  W sEz*+ WE?WeDItO45W"$W"$W'uoW(0~WsCx.W WWWWz6: &.2%)8<*:*.0@8H3;3;3;eCsCx01 2 3;  3; sEz*+ 3;E?3;eDItO453;"$3;"$3;'uo3;(0~3; 3;3;3;3;jZKu---//.-/6: V*V*V*$sCx.)V*eCsCx01 2 V*  V*V*V*V*x.2%)8<*:*.0@8HD"D"D" 3c3h' (D" D" sEz*+ D" E? D"eDItO45D""$D""$D"'uoD"(0~D" D"D"D"D"LEEdEtEEEE2Z  r       ?C & & &5c? &t &t & & & &%4S>%%%%";tCH~;$;;;;604+/$     ###""JNU;U;E#uS#X"678U; tU{ U;U;U;U;n#%#%#%#%#%JZ  3     Z  3       rErIc(eZdZddZdZdZdZdS) rLTrcdS)NTr)rRblockingtimeouts rCacquirez_DummyRLock.acquirestrEcdSrrrs rCreleasez_DummyRLock.release rEc|Srrrs rC __enter__z_DummyRLock.__enter__s rEcdSrr)rRexc_type exc_value tracebacks rC__exit__z_DummyRLock.__exit__rrEN)Tr)rrrrrrr rrErCrLrLsU        rErLr)0loggingrOtypingrrrrrrr numpyr:rkray._common.deprecationr r r ray.rllib.utilsr rray.rllib.utils.frameworkrrray.rllib.utils.metrics.statsrrrrrrrrrrray.util.annotationsrr_tftorch getLoggerloggerr<rjrrDrIrLrrErCrsi @@@@@@@@@@@@@@@@@@ UUUUUUUUUU44444444EEEEEEEE                        98888888 =??2q    q  ; ' '     $# "  BB4S>BDBYBBBBJ Wx  x  x  x  x  x  x  x  v           rE