FigdZddlmZddlZddlZddlZddlZddlZddlZddl m Z ddl m Z ddl m Z ddlZddlZddlmZmZmZddlmZmZmZmZmZmZdd lmZdd lmZdd l m!Z!Gd d Z"y)a5 Module provides functionalities for hyperparameter tuning of the Ultralytics YOLO models for object detection, instance segmentation, image classification, pose estimation, and multi-object tracking. Hyperparameter tuning is the process of systematically searching for the optimal set of hyperparameters that yield the best model performance. This is particularly crucial in deep learning models like YOLO, where small changes in hyperparameters can lead to significant differences in model accuracy and efficiency. Examples: Tune hyperparameters for YOLO26n on COCO8 at imgsz=640 and epochs=10 for 300 tuning iterations. >>> from ultralytics import YOLO >>> model = YOLO("yolo26n.pt") >>> model.tune(data="coco8.yaml", epochs=10, iterations=300, optimizer="AdamW", plots=False, save=False, val=False) ) annotationsN)Counter)datetime)Path) CFG_INT_KEYSget_cfg get_save_dir) DEFAULT_CFGLOGGERYAML callbackscolorstrremove_colorstr)check_requirements) torch_load)plot_tune_resultsceZdZdZedfddZdddZddZdddZe dZ d ddZ dd Z d Z dd Zddd Zdd Ze ddZe d dZe d!d"dZ d# d$dZd%d&dZy)'TuneraSA class for hyperparameter tuning of YOLO models. The class evolves YOLO model hyperparameters over a given number of iterations by mutating them according to the search space and retraining the model to evaluate their performance. Supports both local NDJSON storage and distributed MongoDB Atlas coordination for multi-machine hyperparameter optimization. Attributes: space (dict[str, tuple]): Hyperparameter search space containing bounds and scaling factors for mutation. tune_dir (Path): Directory where evolution logs and results will be saved. tune_file (Path): Path to the NDJSON file where evolution logs are saved. args (SimpleNamespace): Configuration arguments for the tuning process. callbacks (dict): Callback functions to be executed during tuning. prefix (str): Prefix string for logging messages. mongodb (MongoClient): Optional MongoDB client for distributed tuning. collection (Collection): MongoDB collection for storing tuning results. Methods: _mutate: Mutate hyperparameters based on bounds and scaling factors. __call__: Execute the hyperparameter evolution across multiple iterations. Examples: Tune hyperparameters for YOLO26n on COCO8 at imgsz=640 and epochs=10 for 300 tuning iterations. >>> from ultralytics import YOLO >>> model = YOLO("yolo26n.pt") >>> model.tune( >>> data="coco8.yaml", >>> epochs=10, >>> iterations=300, >>> plots=False, >>> save=False, >>> val=False >>> ) Tune with distributed MongoDB Atlas coordination across multiple machines: >>> model.tune( >>> data="coco8.yaml", >>> epochs=10, >>> iterations=300, >>> mongodb_uri="mongodb+srv://user:pass@cluster.mongodb.net/", >>> mongodb_db="ultralytics", >>> mongodb_collection="tune_results" >>> ) Tune with custom search space: >>> model.tune(space={"lr0": (1e-5, 1e-2), "momentum": (0.7, 0.98)}) Nc |jddxsAiddddddd d d d d dddddddddddddddddddddd d!d dddddddd!d" |_|jd#d}|jd$d%}|jd&d'}t|(|_|jj|j_t |j|jjxsd)*|_d+\|j_|j_|j_|jd,z |_ |xstj|_ td-|_ tj|d|_|r|j!|||t#j$|jd.|jd/|jd0y)1zInitialize the Tuner with configurations. Args: args (dict): Configuration for hyperparameter evolution. _callbacks (dict | None, optional): Callback functions to be executed during tuning. spaceNlr0)gh㈵>{Gz?lrf)r?momentum)gffffff?g\(\?g333333? weight_decay)gMbP? warmup_epochs)rg@warmup_momentum)rgffffff?box)rg4@cls)皙?@cls_pw)rrdfl)g?g(@hsv_h)rr"hsv_s)rg?hsv_vdegrees)rgF@ translatescaleshear)rg$@) perspectiveflipudfliplrbgrmosaicmixupcutmix copy_paste close_mosaic mongodb_uri mongodb_db ultralyticsmongodb_collection tuner_results) overridestunename)NFFztune_results.ndjsonzTuner: z*Initialized Tuner instance with 'tune_dir=z' uT💡 Learn about tuning at https://docs.ultralytics.com/guides/hyperparameter-tuning)poprrargsresumeexist_okr r>tune_dir tune_filer get_default_callbacksrprefixadd_integration_callbacksmongodb _init_mongodbr info)selfr@ _callbacksr6r7r9s Z/var/www/pod-logistic/pod-ai/venv/lib/python3.12/site-packages/ultralytics/engine/tuner.py__init__zTuner.__init__Ws6XXgt, 1 <1  ;1  ( 1 L 1 Z 1  {1  ;1  :1  j1  ;1  Z1  Z1  Z1  {1 !1 " [#1 $ [%1 &(    $'71  :hh}d3 XXlM: !XX&:OLd+ !YY-- $TYYTYY^^5MvN ?S<  *DII,<)>>#Hy'F'F'Hy) ++D1    {J8J K {{mEdmm_TW{{mo q c tdddlm}ddlm}m}t |D]Z} ||dddddd d d }|jjd tj|jd|dzd|cSy#||f$rQ||dz k(rd|z}tj|jd|dzd|dtj|YwxYw)afCreate MongoDB client with exponential backoff retry on connection failures. Args: uri (str): MongoDB connection string with credentials and cluster information. max_retries (int): Maximum number of connection attempts before giving up. Returns: (MongoClient): Connected MongoDB client instance. pymongor) MongoClient)ConnectionFailureServerSelectionTimeoutErrori0ui Ni@Ti`)serverSelectionTimeoutMSconnectTimeoutMSsocketTimeoutMS retryWrites retryReads maxPoolSize minPoolSize maxIdleTimeMSpingz$Connected to MongoDB Atlas (attempt )z#MongoDB connection failed (attempt z), retrying in zs...N)rrQrRpymongo.errorsrSrTrangeadmincommandr rJrFwarningtimesleep) rKuri max_retriesrRrSrTattemptclient wait_times rM_connectzTuner._connects 9%'Q[) &G &$-2%*$) $# " !"'  $$V, t{{m+OPWZ[P[}\]^_  & &'BC &kAo-wJ {{m#FwQR{mSbclbmmqr 9% &sABACCc|j||_|j|||_|jjdgdt j |j dy)aInitialize MongoDB connection for distributed tuning. Connects to MongoDB Atlas for distributed hyperparameter optimization across multiple machines. Each worker saves results to a shared collection and reads the latest best hyperparameters from all workers for evolution. Args: mongodb_uri (str): MongoDB connection string, e.g. 'mongodb+srv://username:password@cluster.mongodb.net/'. mongodb_db (str, optional): Database name. mongodb_collection (str, optional): Collection name. Notes: - Creates a fitness index for fast queries of top results - Falls back to local NDJSON mode if connection fails - Uses connection pooling and retry logic for production reliability )fitnessT) backgroundz*Using MongoDB Atlas for distributed tuningN)rorH collection create_indexr rJrF)rKr6r7r9s rMrIzTuner._init_mongodbs\ }}[1 ,,z23EF $$o%64$H t{{m#MNOrOc t|jjjddj |S#t $rgcYSwxYw)zGet top N results from MongoDB sorted by fitness. Args: n (int): Number of top results to retrieve. Returns: (list[dict]): List of result documents with fitness scores and hyperparameters. rqrr)listrtfindsortlimit Exception)rKns rM_get_mongodb_resultszTuner._get_mongodb_resultssK ,,.33IrBHHKL L I sAA AAcPt|dr|jSt|S)z2Convert tensor-like values for JSON serialization.item)hasattrrstr)xs rM _json_defaultzTuner._json_defaults"#1f-qvvx93q69rOc6|t|d||d}|r||d<|S)z%Build one local tuning result record.) iterationrqhyperparametersdatasets save_dirs)round)rKrrqrrrresults rM_result_recordzTuner._result_records2#Wa(.   "+F;  rOc j |jj||jDcic]$\}}|t|dr|j n|&c}}||t j |dycc}}w#t$r.}tj|jd|Yd}~yd}~wwxYw)aSave results to MongoDB with proper type conversion. Args: fitness (float): Fitness score achieved with these hyperparameters. hyperparameters (dict[str, float]): Dictionary of hyperparameter values. metrics (dict): Complete training metrics dictionary (mAP, precision, recall, losses, etc.). datasets (dict[str, dict]): Per-dataset metrics for the iteration. iteration (int): Current iteration number. r)rqrmetricsr timestamprzMongoDB save failed: N) rt insert_oneitemsrrrnowr{r rgrF) rKrqrrrrkves rM_save_to_mongodbzTuner._save_to_mongodbs" E OO & &&_n_t_t_v'wW[WXZ[68JAFFHPQ(Q'w& (!)!*  (x E NNdkk]*?sC D D Es()A;)A5  A;5A;; B2$B--B2cN t|jjjdd}|syt |j dd5}|D]}|j tj|j|d|dxsd|jd i|jd i|jd |j d z dddy#1swYyxYw#t$r.}tj|jd|Yd}~yd}~wwxYw)aSync MongoDB results to the local NDJSON tuning log. Downloads all results from MongoDB and writes them to the local NDJSON file in chronological order. This keeps resume, mutation, and plotting on the same local source of truth when using distributed tuning. rr`Nwutf-8encodingrqrrrrdefault zMongoDB to NDJSON sync failed: )rwrtrxryopenrDwritejsondumpsrgetrr{r rgrF)rK all_resultsfrrs rM_sync_mongodb_to_filezTuner._sync_mongodb_to_files  Ot335::;JKKdnncG< ) FGG  // &{ 3 &y 1 8S & +)weightsrr`rr")minrrisfiniteallsum ones_likerandomchoicesrdstackmaxwhereuniformshape) ralpharridxsi parents_matlohispans rM _crossoverzTuner._crossoverds& 3q6NAqD'AadGKKM)D0{{7#'')W[[]a-?ll7+G~~eCFmWBhh$7Q!QR7; #[__Q%7BBwxx 299#4#4T3 #KTRyy  edl!2B4EFF 8sE/cd}|jr |j|x}rtj|Dcgc]]}|dg|jj Dcgc]0}|dj ||jj |2c}z_c}}}n|jj|jjjvrTtjdg|jj Dcgc]}t|j|c}zg}|!|j|j|}|tjj!t#t%j$t'|j}|j)|} tj|jj+D cgc]} t'| dk(r| dndc} } tj,|} tj.| d k(rtjj||k} tjj1||| zz}tj2| tj4|dj7d d } tj.| d k(rt9|jj Dcic]\}}|t;| || |z}}}n<|jj Dcic]}|t|j|}}|jj=D]1\}}t?tAtC|||d |d d ||<3d|vrt?|d|d<d|vrt?|d|d<|Scc}wcc}}wcc}wcc} wcc}}wcc}w)aMutate hyperparameters based on bounds and scaling factors specified in `self.space`. Args: n (int): Number of top parents to consider. mutation (float): Probability of a parameter mutation in any given iteration. sigma (float): Standard deviation for Gaussian random number generator. Returns: (dict[str, float]): A dictionary containing mutated hyperparameters. Nrqrr)r|rVrbrr`g?r#rrr5epochs)"rHr}rrrkeysrr@rtr>databaselist_collection_namesrrrrseedintrhrrronesrrandnrexpclip enumeraterrrrr)rKr|mutationsigmarrrrnggenesrgainsfactorsmasksteprhypboundss rM_mutatez Tuner._mutatets:  <<33A66w6HH")9aeakakapapar)s\]!,=*>*B*B1diimmTUFV*W)ss %%)A)A)W)W)YYHHsedjjooFW&Xwtyy!'<&XXYZ 9,,T-E-E-G1,MA = IINN3tyy{+ ,TZZBOOA&EHHDJJDUDUDWXqc!fkads:XYEggbkG&&A&yy''+h6yyr*eem<((4s;@@sK&&A&@IIZ?[\tq!1eE!Hwqz122\C\59ZZ__5FG1gdii++GCG))+ FIAv3s3q66!95vayA1ECF F S "'N(;"|dj@}||||dz j rdndz }tC|d}|}|dk(sJdtjDdtGjHtJjLjO|xsd!d"i||<|jU||jWd!xsd"0tY|t|z }|j[|dz|| |t3| |Dcic]\}}|t5|c}}}d#}|jrK|j]|| |||dz|j |j^jai} | |k\rd}n|jc||j}!|je|!}"|"d d df}|jg}#|!|#}$|$jWd$i}%|#|k(}&|&r|r?|jiD],}||%jivstkjl|d%.t3| |D]s\}}'t| dk(r|jdz n|jdz |z }(|(jdd|'jod&D]})tkjp|)|(u|%}n"|r |D]}tkjl|d%|%}tst5|j |j|dzd|d'tj|z d(d)|jd*tud+|jd|jd,||#d-|#dzd|jd.|jw|$d|jd/t|$jWd0idk(r|jdz nd1}*tjd|*zt3|jxj{|"|#dd fD+,cic]$\}+},|+|+t|vr t|,n t|,&} }+},tj|jd2z | t|*j|jd3dz4tj|jd2z |stj|jd5|d6 d7y y cc} wcc}w#tP$r)}tjRd|dzd|Yd }~ad }~wwxYwcc}}wcc},}+w)8aExecute the hyperparameter evolution process when the Tuner instance is called. This method iterates through the specified number of iterations, performing the following steps: 1. Sync MongoDB results to local NDJSON (if using distributed mode) 2. Mutate hyperparameters using the best previous results or defaults 3. Train a YOLO model with the mutated hyperparameters 4. Log fitness scores and hyperparameters to MongoDB and/or NDJSON 5. Track the best performing configuration across all iterations Args: iterations (int): The number of generations to run the evolution for. cleanup (bool): Whether to delete iteration weights to reduce storage space during tuning. T)parentsrBrrzResuming tuning run z from iteration r`z...gr@r皙?r")rzStarting iteration /z with hyperparameters: rr=save_dirsysz-mzultralytics.cfg.__init__trainc30K|]\}}|d|yw)=N).0rrs rM z!Tuner.__call__..s.Yda!AaSz.Ys)checkzbest.ptzlast.pt train_metricsztraining failedz5training failure for hyperparameter tuning iteration rNrqrFr) ignore_errorsz*.ptu iterations complete ✅ (z.2fzs) zResults saved to boldz Best fitness=z observed at iteration zBest fitness metrics are zBest fitness model is rz"not saved for multi-dataset tuningzbest_hyperparameters.yamlz# )rheaderzTarget iterations (z) reached in MongoDB (z ). Stopping.)FrhrCmkdirrHrrDrrrr rJrFrdrrvarsr@r? isinstancerwtuplerr rrzipr __import__ executabler subprocessrun returncoderrigccollecttorchcuda empty_cacher{errorrrrrrrtcount_documentsrrargmaxrshutilrmtreeglobcopy2rrrrrrrrr saverreplaceprint)-rK iterationscleanupt0best_save_dirsstartrfracsigma_i mutated_hyp train_argsr dataset_namesr>rs weights_dirr all_fitnessdataset_metricsjrdataset metrics_ilaunchcmd return_code ckpt_filerrqrstop_after_iterationtotal_mongo_iterationsrrbest_idx best_resultcurrent_best_save_dirsbest_is_current weight_dirbest_weights_dirckptr rrs- rM__call__zTuner.__call__snYY[ D48  "))$)F <<  & & ( >> "0023E KK4;;-';DMM?JZ[`cd[dZeehi juj)x Aq5y#&DC$J&G,,W,5K KK4;;-':1q5':,Nefqers t;DO;{;J>>&)DdT5M2v //5Mt9>gj123O\]tl7:#6TB]  3;;Q1y=;K;GK O#,S}-E#F S->>+JJqMJJLJJ**, ,5+HC8H(""?7#;#?#? #J#QcR9 S:+&[)99G((A36}h3OPZWa#a&P F$) ||%%g{G_VWZ[V[\**,)-)H)H)L&)Z7+/(''/..0G,,W5A1gG~~'H!(+K%0__["%E "&!mO+224A$:$A$A$CC"MM!4@A,/}k+J='GZ58Y!^ 1YbIbelIl%%**4$*G * 7= T+;<= ="8!9AMM!489!7 c$..1 2;;-Awa |3Mdiik\^N^_bMccg;;-0&$--1P0QQS;;-}WX->,??VW_bcWcVddf;;-89K9KK9X8YY[;;-503KOOJPR4S0TXY0Y4==9,`DE F  KKv &MPQUQ[Q[Q`Q`QbdefnpqprfrdsMtuTQAl!2s1va@uDu II ;;&v~~dkk4'HIDP  JJt}}'BB C# {{m#6zlBXYoXpp|}qx "^;6!gLL#XYZ]^Y^X__abcad!effgQfvs1;\. \3C/\81]-%)]38 ]*]%%]*)rL dict | None)z4mongodb+srv://username:password@cluster.mongodb.net/rV)rjrrkr)rCrC)r)r|rreturnrw)N) rrrqrrdict[str, float]rdict[str, dict]rzdict[str, str] | NonerDdict) rqrrrErrGrrFrr)rD list[dict])rrHr|z int | NonerDznp.ndarray | None)rrG)rrGrDrB)rrwrDz list[str])r )r np.ndarrayrrrrrDrJ)rIg?r)r|rrrrrrDrE) T)r#rr$bool)__name__ __module__ __qualname____doc__r rNrorIr} staticmethodrrrrrrrrrrrrArrOrMrr'sP-^(49 v&&PP* ::,0 *  "  )  &EE*E E " E  E>O<D"K  G G" @ @@ @  @DSrOr)#rP __future__rrrrrrrh collectionsrrpathlibrnumpyrrultralytics.cfgrrr ultralytics.utilsr r r r rrultralytics.utils.checksrultralytics.utils.patchesrultralytics.utils.plottingrrrrOrMr[sQ #   ??]]708bbrO