&`izjddlZddlZddlZddlZddlmZddlmZmZm Z m Z m Z m Z m Z mZmZddlZddlmZddlZddlZddlmZddlmZddlmZmZddlm Z m!Z!ddl"m#Z#dd lm$Z$m%Z%m&Z&m'Z'm(Z(m)Z)dd l*m+Z+dd l,m-Z-dd l.m/Z/dd l0m1Z1m2Z2ddl3m4Z4ddl5m6Z6er ddl7m8Z8ddl9m:Z:dZ;dZdZ?ej@eAZBeeCe ee/fZDeeDdfZEGddZFdS)N)Path) TYPE_CHECKINGAnyCallableDictListOptionalTupleTypeUnion)URI) AirEntrypoint)StorageContextget_fs_and_path)V2_MIGRATION_GUIDE_MESSAGE_v2_migration_warnings_enabled)_log_deprecation_warning) ExperimentExperimentAnalysis ResumeConfig RunConfig TuneConfig TuneError)is_function_trainable) ResultGrid) Trainable)_Configrun) flatten_dict)inspect_serializability BaseTrainer)Queuez tuner.pkl _trainable_converted_trainable _param_space_experiment_analysisr"c4eZdZdZddddddddejf dedeej j dee dee dee eefdeed eed ee d efd Zd efdZd eddfdZdZdZ d3dedeefdZdedeefdZde eefdeeefdZdee eefdeeefdZde eefd eeeeeeffdZdede dee eefde deej j f d Zd!eed"d#dee eefd efd$Zd4d%Z e!d e fd&Z"e!d efd'Z#e"j$de fd(Z"e!d ee eeffd)Z%e%j$dee eeffd*Z%de d efd+Z&d e'fd,Z(d e'fd-Z)ded e eeffd.Z*dedee eefd e+fd/Z,dedee eefd e+fd0Z-d1Z.d2Z/dS)5 TunerInternalaOThe real implementation behind external facing ``Tuner``. The external facing ``Tuner`` multiplexes between local Tuner and remote Tuner depending on whether in Ray client mode. In Ray client mode, external ``Tuner`` wraps ``TunerInternal`` into a remote actor, which is guaranteed to be placed on head node. ``TunerInternal`` can be constructed from fresh, in which case, ``trainable`` needs to be provided, together with optional ``param_space``, ``tune_config`` and ``run_config``. It can also be restored from a previous failed run (given ``restore_path``). Args: restore_path: The path from where the Tuner can be restored. If provided, None of the rest args are needed. resume_config: Resume config to configure which trials to continue. trainable: The trainable to be tuned. param_space: Search space of the tuning job. One thing to note is that both preprocessor and dataset can be tuned here. tune_config: Tuning algorithm specific configs. Refer to ray.tune.tune_config.TuneConfig for more info. run_config: Runtime configuration that is specific to individual trials. If passed, this will overwrite the run config passed to the Trainer, if applicable. Refer to ray.tune.RunConfig for more info. N restore_pathstorage_filesystem resume_config trainable param_space tune_config run_config _tuner_kwargs _entrypointc 0ddlm} t|| r=trt dt ||||}|p t|_tj |p t|_ | |_ |r| |||||dS|std|j tjkrIt|j t"jj s%trt dt ||_|jsJ||j||_d|_d|_tj|pi|_d|_|j jpt;j|j|j _t;|j j|j j|j j } | j } | !| j"| #tI| j"tJ&5} | 'tQj)|*ddddS#1swxYwYdS) Nrr!zThe Ray Train + Ray Tune integration has been reworked. Passing a Trainer to the Tuner is deprecated and will be removed in a future release. )tuner_run_configtrainerr.) path_or_urir-overwrite_param_spacer,r+z(You need to provide a trainable to tune.zsThe `RunConfig` class should be imported from `ray.tune` when passing it to the Tuner. Please update your imports. F) storage_pathexperiment_dir_namer+)+ray.train.trainerr" isinstancerrr_choose_run_configr _tune_configcopyr _run_configr2_restore_from_path_or_urirrTUNERraytuner-converted_trainable_validate_trainabler._resume_config _is_restoreddeepcopyr1r'namerget_experiment_dir_namer8r+ create_direxperiment_fs_pathopen_output_streamr _TUNER_PKLas_posixwritepickledumps __getstate__)selfr*r+r,r-r.r/r0r1r2r"storagefsfs p/home/jaya/work/projects/VOICE-AGENT/VIET/agent-env/lib/python3.11/site-packages/ray/tune/impl/tuner_internal.py__init__zTunerInternal.__init__\s 211111 i - - -// (424400!+!'1J (7:<<9Z00?IKK&    * *(#&1+#5 +    F HFGG G  }2 2 2:  ch0< < 2.// (4144 #''''   !9:::&"!!]=99?R$(!   ! P5d6NOO !)6 $ 0 5#/B     ' g0111  " " +Z 8 8 A A C C   7  GGFL!2!2!4!455 6 6 6 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7s:J  JJreturnc|jSN)r?rTs rXget_run_configzTunerInternal.get_run_configs  string_queuer#c(||_||jd<dSN_remote_string_queue)r?r1)rTr0r`s rX&set_run_config_and_remote_string_queuez4TunerInternal.set_run_config_and_remote_string_queues &5A1222r_c<|jdddSrb)r1popr]s rXclear_remote_string_queuez'TunerInternal.clear_remote_string_queues" 5t<<<< > > >$<<>> $$ C..a ' ('  F#6yAA 4 4 4P#PP5LPPP  5 4s A1A A,,A1old_trainable_namec ddlm}||_|jsJ||j|t |j|rF|j}|jtkrt d|j |_dSdS)Nrr!)r-rtzThe Tune experiment will restore using the original run's `RunConfig`. If you made any changes to the `RunConfig` within the Trainer you passed into `Tuner.restore`, they will be ignored in the resumed run.) ray.train.base_trainerr"r-rDrEr;r0rloggerwarningr?)rTr-rr"r5s rX_set_trainable_on_restorez'TunerInternal._set_trainable_on_restores 766666"''''   .$6 !   dnk 2 2 2$(>G!Y[[00?"&!1G    2 2r_new_param_spaceflattened_param_space_keysc|dStt|}||krtd|d|dS)aGDetermines whether the (optionally) re-specified `param_space` is valid. This method performs very loose validation on the new param_space to prevent users from trying to specify new hyperparameters to tune over. Raises: ValueError: if not all keys match the original param_space. NaZInvalid `param_space` input to `Tuner.restore()`. To fix this error, pass in the same `param_space` that was used to initialize the Tuner. Only re-specify the `param_space` to refresh Ray object references that no longer exist due to restoring from a new Ray cluster session. It should not be used to introduce new hyperparameters to tune. Got: z Expected: )sortedrkeysr|)rTrrrs rX _validate_param_space_on_restorez.TunerInternal._validate_param_space_on_restores & - Fl?3388::;; - - -K ! KK /I KK  . -r_c^||_|j||j|dSdS)N)rr)r.r)rTr.rs rX_set_param_space_on_restorez)TunerInternal._set_param_space_on_restore8sM '   '  1 1 $ 0+E 2      ( 'r_ tuner_statec|dd}|dd}||||fS)aLoads Tuner state from the previously saved `tuner.pkl`. Args: tuner_pkl_path: pathlib.Path of the `tuner.pkl` file saved during the original Tuner initialization. Returns: tuple: of `(old_trainable_name, flattened_param_space_keys)` used for validating the re-specified `trainable` and `param_space`. __trainable_nameN__flattened_param_space_keys)rf __setstate__)rTrrrs rX_load_tuner_statezTunerInternal._load_tuner_stateFsT)__-?FF%0__ *D& & " +&&&!#===r_r6r7ct||\}}|t|t5}t j|} dddn #1swxYwY|| \} } | || | || t|} | j |j _ t| j|j _||j _ t%||jj|jj||_n#t.$r d|_YnwxYw||_d|_dS)N)r-r)r.r)experiment_checkpoint_pathdefault_metric default_moder+T)ropen_input_filerrNrOrQloadsreadallrrrr rIr?strparentr8r+rr=metricmoder' ExceptionrFrG) rTr6r-r7r,r+rVfs_pathrWrrrpath_or_uri_objs rXr@z'TunerInternal._restore_from_path_or_uri]s&k3EFF G   Wj 9 9 B B D D E E 4 ,qyy{{33K 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4:>9O9O : : 66 &&4F '    ((-'A )   k** !0 4(+O,B(C(C%.@+ -(:+6#07!.3#5 )))D % %  - - -(,D % % % -, s$'BBB,E EEr4r5r"c|rd|vrtd|rQ|jtjkr*t d|jjd|S|s|jS|S)aChooses which `RunConfig` to use when multiple can be passed in through a Trainer or the Tuner itself. Args: tuner_run_config: The run config passed into the Tuner constructor. trainer: The Trainer instance to use with Tune, which may have a RunConfig specified by the user. param_space: The param space passed to the Tuner. Raises: ValueError: if the `run_config` is specified as a hyperparameter. r0z`RunConfig` cannot be tuned as part of the `param_space`! Move the run config to be a parameter of the `Tuner`: Tuner(..., run_config=RunConfig(...))z6A `RunConfig` was passed to both the `Tuner` and the `zE`. The run config passed to the `Tuner` is the one that will be used.) r|r0rBtrainrrinfo __class____name__)rTr4r5r.s rXr<z TunerInternal._choose_run_configs$  <;668   $ 2ci6I6I6K6K K K KK<%.<<<    $ #   &% % r_c|jd}t|tjjsdS|j|jd<dS)zConverts ``self._param_space["scaling_config"]`` to a dict. The dict is converted back to a dataclass by the Trainer, after the Tune search specification is resolved. scaling_configN)r&getr;rBr ScalingConfig__dict__r>)rTrs rX_process_scaling_configz%TunerInternal._process_scaling_configsZ*../?@@.#)*ABB  F.<.E.J.J.L.L*+++r_c|jSr\)r$r]s rXr-zTunerInternal.trainables r_c|jSr\)r%r]s rXrDz!TunerInternal.converted_trainables ((r_cH||_|||_dSr\)r$_convert_trainabler%)rTr-s rXr-zTunerInternal.trainables%#$($;$;I$F$F!!!r_c|jSr\)r&r]s rXr.zTunerInternal.param_spaces   r_ct|tr|}t|ts"| t dt |d||_|r|dSdS)Nz=The `param_space` passed to the `Tuner` must be a dict. Got 'z ' instead.)r;rto_dictdictr|typer&r)rTr.s rXr.zTunerInternal.param_spaces k7 + + 0%--//K+t,, 1H6[))666  (  +  ( ( * * * * * + +r_cZddlm}t||r|n|S)zConverts a Trainer to a Tune trainable and saves the converted trainable. If not using a Trainer, this leaves the trainable as is.rr!)r:r"r; as_trainable)rTr-r"s rXrz TunerInternal._convert_trainablesC 211111)[11 I " " $ $ $ r_c|j}tj|j}|js|||}n|||}||_t|jSr\) rDr>rHr.rG _fit_internal _fit_resumer'r)rTr-r.analysiss rXfitzTunerInternal.fitsm, mD$455   @)))[AAHH'' ;??H$,!$3444r_cV|jstdt|jS)NziCan't return results as experiment has not been run, yet. Call `Tuner.fit()` to run the experiment first.)r' RuntimeErrorrr]s rX get_resultszTunerInternal.get_resultss8( B $3444r_ctj|jj}|jr8t |dd}|durt d|jd|durd|_|j9t |dd}|durt d |jd |durd|_nt|rd|_nd|_td id |jj d |jj d |jj d|j jd|j jd|jjd|jjd|jjd|jjjd|ddd|jjjd|jjd|jjd|j jd|j jd|j jd|j jd|j jd|jd|j jS)!zrHr?rcheckpoint_frequencygetattrr|checkpoint_at_endrrr8r+rIr=rrrrrfailure_configrrrrrrmrrrr2r)rTr-rhandle_checkpoint_freqhandle_cp_at_ends rX_get_tune_run_argumentsz%TunerInternal._get_tune_run_arguments s!M$*:*LMM  1 ;&-5t&& "&..!,(=,,,   (4//:;!6  . :&y2NPTUU 5(( M(:MMM"T))7<!3 %Y// ;6;!336:!3   )66 #/BB !&& "''  $++  &00  (44 !&& )8EE 0/ #(% '6@@ #.@@ $,, *77 #'"3"I"I! "+99# $ $0CC% &#'"3"I"I' ((() , $0CC-  r_c i||t|||jj|jj|jj|jj|j}tdi|}| |S)zFitting for a fresh Tuner.)run_or_experimentconfigrj search_alg scheduler log_to_filer) rrr=rjrrr?rr1rrgrTr-r.argsrs rXrzTunerInternal._fit_internal^s  **955  "+" -9,7+5 ,8           &&(((r_c|jsJi||t|||j|jj|jj|j}tdi|}||S)zFitting for a restored Tuner.)rrr,rrr) rFrrr=rrr1rrgrs rXrzTunerInternal._fit_resumets""""  **955  "+""1,7+5      ;;;; &&(((r_c"|j}|d|d<|ddd|td|td}|t d}|t d|rtj|nd|d<|.tt| nd|d<|S)Nr1rcrr) rr>rf_TRAINABLE_KEY_CONVERTED_TRAINABLE_KEY_PARAM_SPACE_KEY_EXPERIMENT_ANALYSIS_KEYrr{rrr)rTstater-r.s rXrSzTunerInternal.__getstate__s ""$$!&!7!!>o o""#94@@@ .$'''II6== ii 0$77  *D1119B KJ )) 4 4 4t  ! & < ,,1133 4 4 4 ,-  r_c|dd|dd|j|dS)Nrr)rfrupdate)rTrs rXrzTunerInternal.__setstate__sH 0$777 $d+++ U#####r_r\)rZN)0r __module__ __qualname____doc__rrArr pyarrowrV FileSystemrTrainableTypeOrTrainerrrrrrYr^rdrgrs TrainableTyperErrrrr rr@r<rpropertyr-rDsetterr.rrrrrrrrrSrrr_rXr)r)?s.<!>B046:04,0*.(,%2%8W7W7W7%WZ%:;W7 - W7 23 W7 d38n- W7j)W7Y'W7 ~W7#W7W7W7W7r     B#B3:BBBB ===LLL$RV++&+AI#++++Z2&2S>> x}htCy11 2>>>>.1!1!*1! (S#X7 1! $ 1! %WZ%:; 1!1!1!1!f) "9-) ) d38n- )  ) ) ) ) V M M M M1X)])))X)G#9GGGG!Xd38n5!!!X!+xS#X'?++++"  ,B  }      5Z 5 5 5 55Z5555R R 4S>R R R R h&5=d38n5M ,&5=d38n5M *($$$$$r_r))Gr>rxloggingrkpathlibrtypingrrrrrr r r r pyarrow.fsrray.cloudpickle cloudpicklerQ ray.trainrBray.tuneray.air._internal.uri_utilsr ray.air._internal.usagerray.train._internal.storagerrray.train.constantsrrray.train.utilsrrrrrrrray.tune.registryrray.tune.result_gridrray.tune.trainabler ray.tune.tunerrray.tune.utilsrray.utilr r:r"ray.util.queuer#rNrrrr getLoggerrrrrrr)rr_rXr s                         ++++++111111GGGGGGGG544444433333++++++((((((&&&&&&&&'''''',,,,,,%------$$$$$$ 1!1  8 $ $c8T)_45 }m;<c $c $c $c $c $c $c $c $c $c $r_