.`i$ UddlZddlZddlmZmZddlmZddlmZddl m Z m Z m Z ddl mZddlmZddlmZmZddlmZmZdd lmZdd lmZmZdd lmZdd lmZdd l m!Z!ddl"m#Z#m$Z$m%Z%m&Z&ddl'm(Z(m)Z)ee*Z+eee,gdfZ-e.dZ/e/e-zZ0GddeZ1de2e0fdZ3Gdde1Z4Gdde1Z5Gdde5e4Z6Gdde4Z7Gdde4Z8e e ze zZ9ee:d<de9d e2e,d!e;ded$e,de2e,fd&Z?Gd'd(Z@dS))N)ABCabstractmethod)Callable) TypeAlias)CounterGauge Histogram)CUDAGraphLogging)SupportsMetricsInfo VllmConfig)KVConnectorLoggingKVConnectorPrometheus) init_logger)STAT_LOGGER_PLUGINS_GROUPload_plugins_by_group) FinishReason)PerfMetricsLogging)unregister_vllm_metrics)CachingMetricsIterationStatsMultiModalCacheStatsSchedulerStats)SpecDecodingLoggingSpecDecodingPromStatLoggerBaseAggregateStatLoggerBasec eZdZdZeddedefdZe ddedzde dzd e dzd efd Z ed Z d Z dedefdZdS)raInterface for logging metrics. API users may define custom loggers that implement this interface. However, note that the `SchedulerStats` and `IterationStats` classes are not considered stable interfaces and may change in future versions. r vllm_config engine_indexcdSN)selfrrs k/home/jaya/work/projects/VOICE-AGENT/VIET/agent-env/lib/python3.11/site-packages/vllm/v1/metrics/loggers.py__init__zStatLoggerBase.__init__/sHKNscheduler_statsiteration_statsmm_cache_stats engine_idxcdSr!r"r#r'r(r)r*s r$recordzStatLoggerBase.record2s  sr&cdSr!r"r#s r$log_engine_initializedz%StatLoggerBase.log_engine_initialized;s&)cr&cdSr!r"r/s r$logzStatLoggerBase.log> r&is_awakelevelcdSr!r")r#r4r5s r$record_sleep_statez!StatLoggerBase.record_sleep_stateAr3r&rNr)__name__ __module__ __qualname____doc__rr intr%rrrr-r0r2r7r"r&r$rr'sKKJKcKKK^K 7;   '$. ($. -t3      ^ ))^)    3 s      r&returnc g}ttD]Z\}}t|trt |t std|d|d||[|S)NzStat logger plugin z+ must be a subclass of StatLoggerBase (got z).) rritems isinstancetype issubclassr TypeErrorappend) factoriesname plugin_classs r$!load_stat_logger_plugin_factoriesrJEs)+I34MNNTTVV ' 'l,-- Z .6 6  :d::'3:::  &&&& r&c>eZdZdZededeefdZdS)rzNAbstract base class for loggers that aggregate across multiple DP engines.rengine_indexescdSr!r"r#rrLs r$r%z AggregateStatLoggerBase.__init__ZsLOCr&N) r:r;r<r=rr listr>r%r"r&r$rrVs@--OJOS OOO^OOOr&c eZdZddedefdZdZdefdZde fd Z d ed e de fd Z e d Z ddedzde dzdedzdefdZdZdZdZdZdS)LoggingStatLoggerrrrc||_||_|tjt |_t|_t|_ t|_ t|_ |jj }t||_d|_|jjjr3t'|jjj|jjj|_d|_d|_d|_d|_|rt9||_dSdS)NF)rr_resettime monotonicrlast_scheduler_statsrprefix_caching_metrics connector_prefix_caching_metricsmm_caching_metricsrspec_decoding_loggingkv_transfer_configr kv_connector_loggingcudagraph_loggingobservability_configcudagraph_metricsr compilation_configcudagraph_modecudagraph_capture_sizeslast_prompt_throughputlast_generation_throughputengine_is_idle aggregated_enable_perf_statsrperf_metrics_logging)r#rrr\s r$r%zLoggingStatLogger.__init___s,(& DN$$%%%$2$4$4!'5&6&6#0>0@0@-"0"2"2%8%:%:"!-@$67I$J$J!!%   0 B %5 3B 3K&&D ".1#14'#  " " $ $ H(:;(G(GD % % % H Hr&cL||_d|_d|_d|_d|_dSr9) last_log_timenum_prompt_tokensnum_generation_tokensnum_corrupted_reqsnum_preemptions)r#nows r$rTzLoggingStatLogger._reset}s2 '(*+"'($%r&r?c$|jjjSr!)rr_enable_mfu_metricsr/s r$rhz$LoggingStatLogger._enable_perf_statss4GGr&r(c|xj|jz c_|xj|jz c_|xj|jz c_|xj|jz c_dSr!)rlrmrnronum_preempted_reqs)r#r(s r$_track_iteration_statsz(LoggingStatLogger._track_iteration_statssd /"CC ""o&KK"" ?#EE  BBr& tracked_statsrpcJ||jz }|dkrdSt||z S)NrS)rkfloat)r#rvrp delta_times r$_get_throughputz!LoggingStatLogger._get_throughputs24--   3]Z/000r&c6d|jS)NzEngine {:03d}: )formatrr/s r$ log_prefixzLoggingStatLogger.log_prefixs ''(9:::r&Nr'r)r*cv|r||||j|j|j|j|j|j|j|j|jx}r|j ||j &|j |j |j |j s||_ |jx}r.|r|j||r|j|dSdS)zLog Stats to standard output.N)rurXobserveprefix_cache_statsconnector_prefix_cache_statsrYspec_decoding_statsr[kv_connector_statsr]r^cudagraph_statsrgrW perf_statsrhrirZ)r#r'r(r)r*rrs r$r-zLoggingStatLogger.records`  9  ' ' 8 8 8  &  ' / /0R S S S;G5==#@2>*22?3VWWW%4%GG! F)112DEEE&2#3?&../NOOO? <,;)-88  >d>U>U>W>W >)11*===  <  # + +N ; ; ; ; ; < +4s:      + 5   1 2 2 2 OOD3 4 4 44: R   E F F F OODAJSP Q Q Q&, D   8 9 9 9 OOD3r%rTboolrhrrurxrzpropertyr}rrr-rrr2r0r"r&r$rQrQ^slHHJHcHHHH<&&&HDHHHHCnCCCC1S1u11111;;X;7; !<!<'$.!<($.!<-t3 !<  !<!<!<!z8AggregatedLoggingStatLogger.__init__..s3E E E &)C!!E E E r&)rT)rLlast_scheduler_stats_dictrQr%rgrNs r$r%z$AggregatedLoggingStatLogger.__init__s] -E E -1-@E E E & ""42"FFFr&cPdt|jS)Nz{} Engines Aggregated: )r|lenrLr/s r$r}z&AggregatedLoggingStatLogger.log_prefix$s!(//D4G0H0HIIIr&r?cdS)NFr"r/s r$rhz.AggregatedLoggingStatLogger._enable_perf_stats(sur&Nrr'r(r)r*c||jvrtd|dSt|||||| ||j|<dSdSNzUnexpected engine_idx: %dr)r*)rLrwarningrQr-rr,s r$r-z"AggregatedLoggingStatLogger.record,s{ T0 0 0 NN6 C C C F    )! !     &9HD *: 6 6 6 ' &r&cNt|_|jD]P}|jxj|jz c_|jxj|jz c_|jxj|jz c_Q|jxjt|jzc_dSr!)rrWrvaluesrrrr)r#rWs r$rz5AggregatedLoggingStatLogger.aggregate_scheduler_stats@s$2$4$4!$($B$I$I$K$K    % 6 6$5  6 6  % 6 6$5  6 6  % 4 4$3  4 4 4 !00C8V4W4WW0000r&c:t|dSr!)rQr2r/s r$r2zAggregatedLoggingStatLogger.logNsd#####r&c|jjjr?tdt |j|jjjdSdS)NzR%d Engines: vllm cache_config_info with initialization after num_gpu_blocks is: %d)rrrrrrrLr/s r$r0z2AggregatedLoggingStatLogger.log_engine_initializedQs\   ( 7  KK.D'(( -<        r&r9)r:r;r<r rOr>r%rr}rrhrrrr-rr2r0r"r&r$rrs  S     JJXJD7; II'$.I($.I-t3 I  IIII( X X X$$$r&rc neZdZdedeededdfdZ ddedzd e dzd e dzd efd Z d Z dZ dS)PerEngineStatLoggerAdapterrrLper_engine_stat_logger_factoryr?NcTi|_||_|D]}||||j|<dSr!)per_engine_stat_loggersrL)r#rrLrrs r$r%z#PerEngineStatLoggerAdapter.__init__\sO (*$,*  L9W9W\::D ( 6 6  r&rr'r(r)r*c||jvrtd|dS|j|||||dSr)rrrr-r,s r$r-z!PerEngineStatLoggerAdapter.recordisg T9 9 9 NN6 C C C F $Z077  )! 8     r&cf|jD]}|dSr!)rrr2r#per_engine_stat_loggers r$r2zPerEngineStatLoggerAdapter.logzs@&*&B&I&I&K&K ) ) " " & & ( ( ( ( ) )r&cf|jD]}|dSr!)rrr0rs r$r0z1PerEngineStatLoggerAdapter.log_engine_initialized~s@&*&B&I&I&K&K < < " " 9 9 ; ; ; ; < PerEngineStatLoggerFactoryr%rrrr-r2r0r"r&r$rr[s  S  )C      "7;   '$. ($. -t3       ")))<<<<|-}t!||*|_.|%d?d@gd>|-}t!||*|_/gdA}|%dBdC||-}t!||*|_0|%dDdE||-}t!||*|_1|%dFdG||-} t!| |*|_2|%dHdI||-}!t!|!|*|_3|%dJdK||-}"t!|"|*|_4|%dLdMtM||-}#t!|#|*|_5|jrgdN}$|%dOdP|$|-}%t!|%|*|_6|%dQdR|$|-}&t!|&|*|_7|%dSdT|$|-}'t!|'|*|_8ni|_6i|_7i|_8d|_9|j:tw|jdUkrtx=dVdW|_>dX|_?dY|_@|j:jA|_B|dZd[d\|j>|j?|j@g|_9dSdS)]Nr model_nameenginec4i|]}|t|gSr")str)rrrs r$rz1PrometheusStatLogger.__init__..s3; ; ; ,/C*c#hh'; ; ; r&zvllm:num_requests_runningz.Number of requests in model execution batches. mostrecentrH documentationmultiprocess_mode labelnameszvllm:num_requests_waitingz+Number of requests waiting to be processed.zvllm:engine_sleep_statezEngine sleep state; awake = 0 means engine is sleeping; awake = 1 means engine is awake; weights_offloaded = 1 means sleep level 1; discard_all = 1 means sleep level 2. sleep_state)rHrrr)awakeweights_offloaded discard_allcBi|]}||S))rrr)labels)rrgauge_engine_sleep_staterss r$rz1PrometheusStatLogger.__init__..sJ000-44:15000r&zvllm:kv_cache_usage_percz*KV-cache usage. 1 means 100 percent usage.zvllm:corrupted_requestszMCorrupted requests, in terms of total number of requests with NaNs in logits.)rHrrzvllm:prefix_cache_queriesz;Prefix cache queries, in terms of number of queried tokens.zvllm:prefix_cache_hitsz7Prefix cache hits, in terms of number of cached tokens.z"vllm:external_prefix_cache_querieszsExternal prefix cache queries from KV connector cross-instance cache sharing, in terms of number of queried tokens.zvllm:external_prefix_cache_hitszoExternal prefix cache hits from KV connector cross-instance cache sharing, in terms of number of cached tokens.zvllm:mm_cache_queriesz?Multi-modal cache queries, in terms of number of queried items.zvllm:mm_cache_hitsz;Multi-modal cache hits, in terms of number of cached items.zvllm:num_preemptionsz0Cumulative number of preemption from the engine.zvllm:prompt_tokensz#Number of prefill tokens processed.zvllm:generation_tokensz&Number of generation tokens processed.zvllm:request_successz)Count of successfully processed requests.finished_reasonc ti|]4}|t|t5Sr"rr)rrcounter_request_success_baserreasons r$rz1PrometheusStatLogger.__init__..asQ444188C#f++444r&zvllm:request_prompt_tokens)rHrbucketsrzvllm:request_generation_tokenszvllm:iteration_tokens_totalz.Histogram of number of tokens per engine_step.)  @iiiii i@z&vllm:request_max_num_generation_tokensz;Histogram of maximum number of requested generation tokens.zvllm:request_params_nz%Histogram of the n request parameter.)r zvllm:request_params_max_tokensz.Histogram of the max_tokens request parameter.z vllm:time_to_first_token_secondsz,Histogram of time to first token in seconds.)MbP?{Gzt?{Gz?{Gz?g{Gz?gQ?g{Gz?皙?g????@@@$@4@D@T@gd@g@g@z vllm:inter_token_latency_secondsz,Histogram of inter-token latency in seconds.)rg?皙?g333333?rg333333?皙?333333?g?rrrrrrrrrrz*vllm:request_time_per_output_token_secondsz7Histogram of time_per_output_token_seconds per request.)rrg?rg?g@rrrg.@rg>@rgI@gN@g^@gn@g~@g@g@g@z vllm:e2e_request_latency_secondsz,Histogram of e2e request latency in seconds.zvllm:request_queue_time_secondsz5Histogram of time spent in WAITING phase for request.z#vllm:request_inference_time_secondsz5Histogram of time spent in RUNNING phase for request.z!vllm:request_prefill_time_secondsz5Histogram of time spent in PREFILL phase for request.z vllm:request_decode_time_secondsz4Histogram of time spent in DECODE phase for request.z'vllm:request_prefill_kv_computed_tokenszMHistogram of new KV tokens computed during prefill (excluding cached tokens).)rgMb`?rrrrrrrrrrrr<xi,iXiizvllm:kv_block_lifetime_secondsz|Histogram of KV cache block lifetime from allocation to eviction. Sampled metrics (controlled by --kv-cache-metrics-sample).z'vllm:kv_block_idle_before_evict_secondszqHistogram of idle time before KV cache block eviction. Sampled metrics (controlled by --kv-cache-metrics-sample).zvllm:kv_block_reuse_gap_secondszHistogram of time gaps between consecutive KV cache block accesses. Only the most recent accesses are recorded (ring buffer). Sampled metrics (controlled by --kv-cache-metrics-sample).rzfvllm:lora_requests_info prometheus metrics may be incorrect/misleading with data parallel deployments.max_lorawaiting_lora_adaptersrunning_lora_adapterszvllm:lora_requests_infozRunning stats on lora requests.sum)CrLrrr_show_hidden_metricskv_cache_metricskv_cache_metrics_enabled model_configserved_model_name max_model_len_spec_decoding_clsspeculative_configspec_decoding_prom_kv_connector_clskv_connector_prom _gauge_clsmake_per_enginegauge_scheduler_runninggauge_scheduler_waitingrr7gauge_kv_cache_usagerr _counter_clscounter_corrupted_requestscounter_prefix_cache_queriescounter_prefix_cache_hits&counter_connector_prefix_cache_queries#counter_connector_prefix_cache_hitscounter_mm_cache_queriescounter_mm_cache_hitscounter_num_preempted_reqscounter_prompt_tokenscounter_generation_tokenscounter_request_successr_histogram_clsbuild_1_2_5_buckets#histogram_num_prompt_tokens_request'histogram_num_generation_tokens_requesthistogram_iteration_tokens+histogram_max_num_generation_tokens_requesthistogram_n_requesthistogram_max_tokens_requesthistogram_time_to_first_tokenhistogram_inter_token_latency'histogram_request_time_per_output_tokenhistogram_e2e_time_requesthistogram_queue_time_request histogram_inference_time_requesthistogram_prefill_time_requesthistogram_decode_time_request%histogram_prefill_kv_computed_requesthistogram_kv_block_lifetime$histogram_kv_block_idle_before_evicthistogram_kv_block_reuse_gapgauge_lora_info lora_configrrrlabelname_max_loralabelname_waiting_lora_adapterslabelname_running_lora_adapters max_lorasr)-r#rrLrrper_engine_labelvaluesr rrrrrrrrrrrrrrrr r!r"r#r$r%r&request_latency_bucketsr'r(r)r*r+r,kv_cache_residency_bucketsr-r.r/rrrrrs- @@@@@r$r%zPrometheusStatLogger.__init__s   !SN,!!!&$/#C#W  , = %#H-  -? #0> ; ; ; ; 3A; ; ; #'"9"9  *J8N# # "&!7!7 %;" " #'//,J*! #2# #  (7 #^Z( ( $#'//,G*! #2# #  (7 #^Z( ( $$(??*7 "]O3*$3 $ $  )+%CCC   A000000* 000D )! , , !!!#+F*! /   %4 .*% % !  + )-):):.+& *;** &/>*NJ//D +(,'8'8,M! (9( ( $-< (.*- - )%)$5$5)T!%6% % ! *9 %~z* * &261B1B5V" 2C2 2 .7F 2NJ7 7 3/3.?.?2U" /@/ / +4C /4 4 0$(#4#4(Q! $5$ $  )8 $nj) ) %!% 1 1%M! !2! ! &5 !>:& & "&*%6%6'L!&7& & " +: & + + '!% 1 1%?!!2! !  &5 !>:& & "%)$5$5)B!%6% % ! *9 %~z* * &PR$'+'8'8'E!%6$77(9( ( $ #  F444444* 444D ( 0 0/3.A.A-?' 66! /B/ / + 4C /4 4 0372E2E1B' 66! 3F3 3 / 8G 3^Z8 8 4&*%8%8.JTTT! &9& & " +: & + + '7;6I6I9W' 66! 7J7 7 3 :. . *)-(;(;3H*"1)<) ) %4.= )>:. . *372E2E=S*"13F3 3 /48G 3^Z8 8 4# # # .&*%8%83H+! &9& & " +: & + + '(,':':2Q+! (;( ( $ -< (.*- - ),0+>+>6Q+! ,?, , ( 1@ ,nj1 1 -*.)<)<4Q+! *=* * & /> *NJ/ / +)-(;(;3P+! )<) ) % .= )>:. . *150C0C:-( 66!1D1 1 -6E 1>:6 6 2  (D 3*** &0+/*=*=5Q3%+>++ '0?+^Z00D ,483F3F>Q3%4G44 09H4nj99D 5,0+>+>62 3%,? , , (1@,nj11D - -02D ,8:D 502D -.2  " .4&''!++K'1D #3JD 03JD 0'3=DM#'??.?"'+88 $3 $ $D  / .r&rC config_objcj|}d|d<d\}}|dkrd}d}| Jd||||d| }|jD]H}|}t ||d<|jd i|d IdS) NrNNrzvllm:cache_config_infoz(Information of the LLMEngine CacheConfigzUnknown metrics info type rrrr") metrics_infor keysrLrrset)r#rCr9r=rHr info_gaugers r$log_metrics_infoz%PrometheusStatLogger.log_metrics_infos!..00 !# X(m > ! !+DFM!Dd!D!D __'*#((** %  !/ 5 5L%2244L%(%6%6L " J  - - - - 1 1! 4 4 4 4 5 5r&rr'r(r)r*c^ ||j||j|j||j|j||j|j||j j |j ||j j |j T|j||j j |j||j j |j |j|j||j |j|j||jr|jr|j|}|j|}|j|}|jD]U}||j||j|jD]} || V|jd|j } d|j! } |j"| |j#| |j$|j%i} |jj&di| '|J|j(||j |j)||j |dStTj+r%|j,||j-|j.||j/|j0||j1|j2||j3|j4||j1|j3z|j5D]"} |j6|| #|j7D]"}|j8||#|j9D]"}|j:||#|j;D]"}|j<||#|j=D]}|j>|j?||j@||jA|jB||jC|jD||jE|jF||jG|jH||jI|j1t|jKdz }|jL|||jM||j1|jN||j3|jO||jP|jQr%|jR||jQdS)zLog to prometheus.N,rr")Sr r?rrrrrrincrqueriesrhitsrrrrrrrr rkv_cache_eviction_eventsr-r.r/lifetime_seconds idle_secondsreuse_gaps_secondsr0rrr>rr4r3r2rrset_to_current_timerrrrrrnrrtrrlrrmr max_num_generation_tokens_iterr! n_params_iterr"time_to_first_tokens_iterr$inter_token_latencies_iterr%finished_requestsr finish_reasonr' e2e_latencyr( queued_timer* prefill_timer)inference_timer+ decode_timemaxnum_cached_tokensr,rrr&mean_time_per_output_tokenmax_tokens_paramr#)r#r'r(r)r* lifetime_hist idle_hist reuse_histeventgaprrlora_info_labelsmax_gen_tokensn_paramttftitlfinished_requestprefill_kv_computeds r$r-zPrometheusStatLogger.records  &  ( 4 8 80     ( 4 8 80     %j 1 5 5o6T U U U  -j 9 = =2:     *: 6 : :27   ;G;JGKK#@H8DHH#@E2>'//#71=&..#6  - 0#< 0!% @ L  EjQ !>zJ ,E00E!))%*@AAA%%e&8999$700"**3////0#/(+#9>>@@))%),#9>>@@))%8:O8:O+T]$ ,$+??.>??SSUUU  %  )* 5 9 9.:P Q Q Q  &z 2 6 6~7J K K K  " F  +   +J 7 ; ;2    ' 377  .    ":.22?3TUUU &z266  1    ' 3;;  -0U U   .L  N  z J N N!     )' 2: > B B5 I I I I  J Jr&cF|d|jjdS)Nr)rArrr/s r$r0z+PrometheusStatLogger.log_engine_initialized~s$ nd.>.KLLLLLr&r!r9rr)r:r;r<rr rrr rrrrr r rOr>r%rr rArrrr-r7r0r"r&r$rrs5JLN)-KOyy%y7;Cy47Gyyyyv5S56I555587; LL'$.L($.L-t3 L  LLLL\JJJJJJJ&MMMMMr&r PromMetricmetric engine_idxsrc$fd|DS)Nc Xi|]&}|t|'Sr"r)rrrlrs r$rz#make_per_engine..s/ L L LCz3s8844 L L Lr&r")rlrmrs` `r$r r s$ M L L L L L L LLr& mantissa_lst max_valuecpd}g} |D])}|d|zz}||kr||&|cS|dz }2)z Builds a list of buckets with increasing powers of 10 multiplied by mantissa values until the value exceeds the specified maximum. rTrr)rF)rprqexponentrmvalues r$ build_bucketsrvsg HG  AH $E !!u%%%%A r&c&tgd|S)zR Example: >>> build_1_2_5_buckets(100) [1, 2, 5, 10, 20, 50, 100] )rrr)rv)rqs r$rrs I . ..r&c eZdZdZ ddedeedzdeedzd ed ed ef d Z dd e dzde dzde dzdedzfdZ ddedefdZdZdZdS)StatLoggerManagera StatLoggerManager: Logging happens at the level of the EngineCore (per scheduler). * DP: >1 EngineCore per AsyncLLM - loggers for each EngineCore. * With Local Logger, just make N copies for N EngineCores. * With Prometheus, we need a single logger with N "labels" This class abstracts away this implementation detail from the AsyncLLM, allowing the AsyncLLM to just call .record() and .log() to a simple interface. NTFrrrmcustom_stat_loggersenable_default_loggersaggregate_engine_logging client_countc|r|ndg|_g|_g}||||rjtt jrF|dkrtdn%|rtnt}| |d} |D]} t| tr?t| tr*| ||j} t| trd} nt!||j| } |j | | s/|j t||jdSdS)NrrzfAsyncLLM created with api_server_count more than 1; disabling stats logging to avoid incomplete stats.F)rrLT)rrLr)rL stat_loggersrr isEnabledForloggingINFOrrrQrFrBrCrDrrr) r#rrmrzr{r|r}stat_logger_factoriesdefault_logger_factorycustom_prometheus_loggerstat_logger_factoryglobal_stat_loggers r$r%zStatLoggerManager.__init__s.9Akkqc;=9;  * ! ( ()< = = = ! Ef&9&9',&G&G EaI0+//*' &,,-CDDD). #8 9 9 -t44 #%<:: &9%8 +#'#6&&&"02FGG4/3,&@ +#'#63F&&&"   $ $%7 8 8 8 8'    $ $$[$2EFF       r&r'r(r)r*cT|d}|jD]}|||||dS)Nrr)rr-)r#r'r(r)r*rs r$r-zStatLoggerManager.recordsW  J'  F MM-%       r&rrgr5cF|jD]}|||dSr!)rr7)r#rgr5rs r$r7z$StatLoggerManager.record_sleep_states7' 4 4F  % %eU 3 3 3 3 4 4r&cB|jD]}|dSr!)rr2)r#rs r$r2zStatLoggerManager.logs-'  F JJLLLL  r&cB|jD]}|dSr!)rr0)r# agg_loggers r$r0z(StatLoggerManager.log_engine_initializeds3+ 0 0J  - - / / / / 0 0r&)NNTFrr<rj)r:r;r<r=r rOr>StatLoggerFactoryrr%rrrr-r7r2r0r"r&r$ryrys6  )->B'+).111#Y%1""34t; 1 !% 1 #' 11111n7;!% '$.($.-t3  $J "444444400000r&ry)ArrUabcrrcollections.abcrtypingrprometheus_clientrrr vllm.envsrvllm.compilation.cuda_graphr vllm.configr r 4vllm.distributed.kv_transfer.kv_connector.v1.metricsr r vllm.loggerr vllm.pluginsrrvllm.v1.enginervllm.v1.metrics.perfrvllm.v1.metrics.prometheusrvllm.v1.metrics.statsrrrrvllm.v1.spec_decode.metricsrrr:rr>rrCAggregateStatLoggerFactoryrrrOrJrrQrrrrk__annotations__objectdictr rvrryr"r&r$rs ########$$$$$$777777777788888877777777$#####IIIIIIII''''''333333>>>>>> NMMMMMMM X  %z3&79I&IJ!";<.1KK     S   <40A+B"PPPPPnPPPvvvvvvvvrAAAAA"35LAAAH%<%<%<%<%