Xc@@s%ddlmZddlmZddlZddlZddlZddlZddlZddl Z ddl m Z ddl m Z ddl m Z ddlmZdd lmZyddlZWnek rdZnXejd krddlZnd efd YZd efdYZdefdYZdefdYZdefdYZdefdYZdefdYZdefdYZ defdYZ!defdYZ"defd YZ#d!efd"YZ$d#efd$YZ%dS(%i(tabsolute_import(tprint_functionN(tdeque(t OrderedDict(tIterablei(tProgbar(tbackendt tensorflowt CallbackListcB@seZdZd ddZdZdZdZd dZd dZ d dZ d d Z d d Z d d Z d ZRS(sContainer abstracting a list of callbacks. # Arguments callbacks: List of `Callback` instances. queue_length: Queue length for keeping running statistics over callback execution time. i cC@s5|p g}g|D] }|^q|_||_dS(N(t callbackst queue_length(tselfR R tc((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyt__init__$s cC@s|jj|dS(N(R tappend(R tcallback((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyR)scC@s%x|jD]}|j|q WdS(N(R t set_params(R tparamsR((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyR,scC@s%x|jD]}|j|q WdS(N(R t set_model(R tmodelR((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyR0scC@sm|p i}x!|jD]}|j||qWd|_tgd|j|_tgd|j|_dS(sCalled at the start of an epoch. # Arguments epoch: integer, index of epoch. logs: dictionary of logs. gtmaxlenN(R ton_epoch_begint_delta_t_batchRR t_delta_ts_batch_begint_delta_ts_batch_end(R tepochtlogsR((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyR4s   cC@s4|p i}x!|jD]}|j||qWdS(sCalled at the end of an epoch. # Arguments epoch: integer, index of epoch. logs: dictionary of logs. N(R t on_epoch_end(R RRR((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyRBs cC@s|p i}tj}x!|jD]}|j||q"W|jjtj|tj|j}|jdkr|d|jkr|dkrtj d|ntj|_ dS(sCalled right before processing a batch. # Arguments batch: integer, index of batch within the current epoch. logs: dictionary of logs. ggffffff?g?sXMethod on_batch_begin() is slow compared to the batch update (%f). Check your callbacks.N( ttimeR ton_batch_beginRRtnptmedianRtwarningstwarnt_t_enter_batch(R tbatchRtt_before_callbacksRtdelta_t_median((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyRMs     cC@s|p i}t|ds-tj|_ntj|j|_tj}x!|jD]}|j||qYW|jjtj|tj |j}|jdkr|d|jkr|dkrt j d|ndS(sCalled at the end of a batch. # Arguments batch: integer, index of batch within the current epoch. logs: dictionary of logs. R"ggffffff?g?sVMethod on_batch_end() is slow compared to the batch update (%f). Check your callbacks.N( thasattrRR"RR t on_batch_endRRRRR R!(R R#RR$RR%((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyR'bs   cC@s1|p i}x|jD]}|j|qWdS(shCalled at the beginning of training. # Arguments logs: dictionary of logs. N(R ton_train_begin(R RR((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyR(xs cC@s1|p i}x|jD]}|j|qWdS(sbCalled at the end of training. # Arguments logs: dictionary of logs. N(R t on_train_end(R RR((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyR)s cC@s t|jS(N(titerR (R ((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyt__iter__sN(t__name__t __module__t__doc__tNoneR RRRRRRR'R(R)R+(((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyRs       tCallbackcB@sqeZdZdZdZdZd dZd dZd dZ d dZ d dZ d d Z RS( sAbstract base class used to build new callbacks. # Properties params: dict. Training parameters (eg. verbosity, batch size, number of epochs...). model: instance of `keras.models.Model`. Reference of the model being trained. The `logs` dictionary that callback methods take as argument will contain keys for quantities relevant to the current batch or epoch. Currently, the `.fit()` method of the `Sequential` model class will include the following quantities in the `logs` that it passes to its callbacks: on_epoch_end: logs include `acc` and `loss`, and optionally include `val_loss` (if validation is enabled in `fit`), and `val_acc` (if validation and accuracy monitoring are enabled). on_batch_begin: logs include `size`, the number of samples in the current batch. on_batch_end: logs include `loss`, and optionally `acc` (if accuracy monitoring is enabled). cC@s d|_dS(N(R/tvalidation_data(R ((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyR scC@s ||_dS(N(R(R R((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyRscC@s ||_dS(N(R(R R((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyRscC@sdS(N((R RR((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyRscC@sdS(N((R RR((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyRscC@sdS(N((R R#R((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyRscC@sdS(N((R R#R((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyR'scC@sdS(N((R R((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyR(scC@sdS(N((R R((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyR)sN( R,R-R.R RRR/RRRR'R(R)(((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyR0s        t BaseLoggercB@s2eZdZddZddZddZRS(s{Callback that accumulates epoch averages of metrics. This callback is automatically applied to every Keras model. cC@sd|_i|_dS(Ni(tseenttotals(R RR((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyRs cC@s|p i}|jdd}|j|7_xT|jD]F\}}||jkro|j|c||7R=ttargetRFi( RFtprintRGRARRHRtprogbarR3(R RRRH((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyRs    cC@s"|j|jkrg|_ndS(N(R3RHt log_values(R R#R((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyR scC@s|p i}|jdd}|jr9|jd7_n|j|7_x>|jdD]/}||krV|jj|||fqVqVW|jr|j|jkr|jj |j|jndS(NR5iiR;( R6RAR3RRKRRFRHRJtupdate(R R#RR8R9((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyR' s   !cC@s||p i}x>|jdD]/}||kr|jj|||fqqW|jrx|jj|j|jdtndS(NR;tforce(RRKRRFRJRLR3RB(R RRR9((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyRs   ! N( R,R-R.R R/R(RRR'R(((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyR<s    tHistorycB@s&eZdZddZddZRS(sCallback that records events into a `History` object. This callback is automatically applied to every Keras model. The `History` object gets returned by the `fit` method of models. cC@sg|_i|_dS(N(Rthistory(R R((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyR(/s cC@sY|p i}|jj|x6|jD](\}}|jj|gj|q)WdS(N(RRR7ROt setdefault(R RRR9R:((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyR3s N(R,R-R.R/R(R(((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyRN's tModelCheckpointcB@s5eZdZddeedddZddZRS(s~Save the model after every epoch. `filepath` can contain named formatting options, which will be filled the value of `epoch` and keys in `logs` (passed in `on_epoch_end`). For example: if `filepath` is `weights.{epoch:02d}-{val_loss:.2f}.hdf5`, then the model checkpoints will be saved with the epoch number and the validation loss in the filename. # Arguments filepath: string, path to save the model file. monitor: quantity to monitor. verbose: verbosity mode, 0 or 1. save_best_only: if `save_best_only=True`, the latest best model according to the quantity monitored will not be overwritten. mode: one of {auto, min, max}. If `save_best_only=True`, the decision to overwrite the current save file is made based on either the maximization or the minimization of the monitored quantity. For `val_acc`, this should be `max`, for `val_loss` this should be `min`, etc. In `auto` mode, the direction is automatically inferred from the name of the monitored quantity. save_weights_only: if True, then only the model's weights will be saved (`model.save_weights(filepath)`), else the full model is saved (`model.save(filepath)`). period: Interval (number of epochs) between checkpoints. tval_lossitautoicC@s#tt|j||_||_||_||_||_||_d|_ |dkr{t j d|t d}n|dkrt j|_t j|_n}|dkrt j|_t j |_nUd|jks|jjdrt j|_t j |_nt j|_t j|_dS( NiRStmintmaxs:ModelCheckpoint mode %s is unknown, fallback to auto mode.tacctfmeasure(RSsminsmax(R?RQR tmonitorRFtfilepathtsave_best_onlytsave_weights_onlytperiodtepochs_since_last_saveR R!tRuntimeWarningRtlesst monitor_optInftbesttgreatert startswith(R RYRXRFRZR[tmodeR\((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyR Zs0              !  cC@s|p i}|jd7_|j|jkrd|_|jjd||}|jrI|j|j}|dkrtj d|jt q|j ||j r|j dkrtd||j|j ||fn||_ |jr|jj|dtqF|jj|dtq|j dkrtd||jfqq|j dkrotd||fn|jr|jj|dtq|jj|dtndS( NiiRs5Can save best model only with %s available, skipping.s?Epoch %05d: %s improved from %0.5f to %0.5f, saving model to %st overwritesEpoch %05d: %s did not improvesEpoch %05d: saving model to %s(R]R\RYtformatRZR6RXR/R R!R^R`RbRFRIR[Rt save_weightsRBtsave(R RRRYtcurrent((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyRzs6        N(R,R-R.R@R R/R(((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyRQ:s t EarlyStoppingcB@sJeZdZddddddZddZddZddZRS( sJStop training when a monitored quantity has stopped improving. # Arguments monitor: quantity to be monitored. min_delta: minimum change in the monitored quantity to qualify as an improvement, i.e. an absolute change of less than min_delta, will count as no improvement. patience: number of epochs with no improvement after which training will be stopped. verbose: verbosity mode. mode: one of {auto, min, max}. In `min` mode, training will stop when the quantity monitored has stopped decreasing; in `max` mode it will stop when the quantity monitored has stopped increasing; in `auto` mode, the direction is automatically inferred from the name of the monitored quantity. RRiRScC@stt|j||_||_||_||_d|_d|_|d krut j d|j t d}n|dkrt j|_nW|dkrt j|_n<d|jks|jjdrt j|_n t j|_|jt jkr |jd9_n|jd 9_dS( NiRSRTRUs8EarlyStopping mode %s is unknown, fallback to auto mode.RVRWii(sautosminsmax(R?RkR RXtpatienceRFt min_deltatwaitt stopped_epochR R!ReR^RR_R`RcRd(R RXRmRlRFRe((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyR s,           ! cC@s5d|_|jtjkr$tjntj |_dS(Ni(RnR`RR_RaRb(R R((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyR(s cC@s|j|j}|dkr8tjd|jtn|j||j|jri||_d|_ n9|j |j kr||_ t |j _n|j d7_ dS(Ns%Early stopping requires %s available!ii(R6RXR/R R!R^R`RmRbRnRlRoRBRt stop_training(R RRRj((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyRs     cC@s6|jdkr2|jdkr2td|jndS(NisEpoch %05d: early stopping(RoRFRI(R R((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyR)sN(R,R-R.R R/R(RR)(((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyRks   t RemoteMonitorcB@s/eZdZdddddZddZRS(sCallback used to stream events to a server. Requires the `requests` library. Events are sent to `root + '/publish/epoch/end/'` by default. Calls are HTTP POST, with a `data` argument which is a JSON-encoded dictionary of event data. # Arguments root: String; root url of the target server. path: String; path relative to `root` to which the events will be sent. field: String; JSON field under which the data will be stored. headers: Dictionary; optional custom HTTP headers. Defaults to: `{'Accept': 'application/json', 'Content-Type': 'application/json'}` shttp://localhost:9000s/publish/epoch/end/tdatacC@s^tt|j|dkr6idd6dd6}n||_||_||_||_dS(Nsapplication/jsontAccepts Content-Type(R?RqR R/troottpathtfieldtheaders(R RtRuRvRw((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyR s      cC@stdkrtdn|p$i}i}||d= 1.0.i(R?RR RXRCtfactortmin_lrtepsilonRlRFtcooldowntcooldown_counterRnRbReR/R`t_reset( R RXRRlRFReRRR((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyR s              c@sjd kr2tjdjtd_njdks_jdkrdjkrfd_tj_nfd_tj _d_ d_ j d _ d S( s2Resets wait counter and cooldown counter. RSRTRUsILearning Rate Plateau Reducing mode %s is unknown, fallback to auto mode.RVc@stj||jS(N(RR_R(tatb(R (s./tmp/pip-build-isqEY4/keras/keras/callbacks.pytsc@stj||jS(N(RRcR(RR(R (s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyRsig-C6?N(sautosminsmax( ReR R!R^RXR`RRaRbRRnRt lr_epsilon(R ((R s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyRs     cC@s|jdS(N(R(R R((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyR(scC@s|p i}tj|jjj|d<|j|j}|dkr`tj d|jt n.|j r|j d8_ d|_ n|j||jr||_d|_ n|j s|j |jkr|ttj|jjj}||j|jkr|||j}t||j}tj|jjj||jdkratd||fn|j|_ d|_ q|n|j d7_ ndS(NRs5Learning Rate Plateau Reducing requires %s available!iis* Epoch %05d: reducing learning rate to %s.(Rt get_valueRRRR6RXR/R R!R^t in_cooldownRRnR`RbRlRRRRRURRFRIR(R RRRjtold_lrtnew_lr((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyRs0          cC@s |jdkS(Ni(R(R ((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyR sN( R,R-R.R RR/R(RR(((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyRs"    t CSVLoggercB@sAeZdZdedZddZddZddZRS(sTCallback that streams epoch results to a csv file. Supports all values that can be represented as a string, including 1D iterables such as np.ndarray. # Example ```python csv_logger = CSVLogger('training.log') model.fit(X_train, Y_train, callbacks=[csv_logger]) ``` # Arguments filename: filename of the csv file, e.g. 'run/log.csv'. separator: string used to separate elements in the csv file. append: True: append if file exists (useful for continuing training). False: overwrite existing file, t,cC@sM||_||_||_d|_d|_t|_tt |j dS(N( tseptfilenameRR/RtkeysRBt append_headerR?RR (R Rt separatorR((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyR !s      cC@s|jrmtjj|jrUt|j#}tt|j |_ WdQXnt|jd|_ nt|jd|_ dS(NRtw( RtosRutexistsRtopentboolRtreadlineRtcsv_file(R Rtf((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyR(*s  %c@sp idjstj_dtjffdY}tjjddgjd|_jrjjqnt i|d6}|j fdjDjj |jj dS(NcS@sYt|tjo|jdk}t|trQ| rQddjtt|S|SdS(Nis"[%s]"s, (RRtndarraytndimRtjointmapRD(R9tis_zero_dim_ndarray((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyt handle_value6s!t CustomDialectc@seZjZRS((R,R-Rt delimiter((R (s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyR@st fieldnamesRtdialectc3@s%|]}||fVqdS(N((t.0tkey(RR(s./tmp/pip-build-isqEY4/keras/keras/callbacks.pys Is( RtsortedRtcsvtexcelt DictWriterRRt writeheaderRRLtwriterowR(R RRRtrow_dict((RRR s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyR3s    #cC@s|jjd|_dS(N(RRR/R(R R((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyR)Ms N( R,R-R.R@R R/R(RR)(((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyRs  tLambdaCallbackcB@s)eZdZdddddddZRS(s1Callback for creating simple, custom callbacks on-the-fly. This callback is constructed with anonymous functions that will be called at the appropriate time. Note that the callbacks expects positional arguments, as: - `on_epoch_begin` and `on_epoch_end` expect two positional arguments: `epoch`, `logs` - `on_batch_begin` and `on_batch_end` expect two positional arguments: `batch`, `logs` - `on_train_begin` and `on_train_end` expect one positional argument: `logs` # Arguments on_epoch_begin: called at the beginning of every epoch. on_epoch_end: called at the end of every epoch. on_batch_begin: called at the beginning of every batch. on_batch_end: called at the end of every batch. on_train_begin: called at the beginning of model training. on_train_end: called at the end of model training. # Example ```python # Print the batch number at the beginning of every batch. batch_print_callback = LambdaCallback( on_batch_begin=lambda batch,logs: print(batch)) # Plot the loss after every epoch. import numpy as np import matplotlib.pyplot as plt plot_loss_callback = LambdaCallback( on_epoch_end=lambda epoch, logs: plt.plot(np.arange(epoch), logs['loss'])) # Terminate some processes after having finished model training. processes = ... cleanup_callback = LambdaCallback( on_train_end=lambda logs: [ p.terminate() for p in processes if p.is_alive()]) model.fit(..., callbacks=[batch_print_callback, plot_loss_callback, cleanup_callback]) ``` cK@stt|j|jj||dk r;||_n d|_|dk r_||_n d|_|dk r||_n d|_|dk r||_ n d|_ |dk r||_ n d|_ |dk r||_ n d|_ dS(NcS@sdS(N(R/(RR((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyRscS@sdS(N(R/(RR((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyRscS@sdS(N(R/(R#R((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyRscS@sdS(N(R/(R#R((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyRscS@sdS(N(R/(R((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyRscS@sdS(N(R/(R((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyRs( R?RR t__dict__RLR/RRRR'R(R)(R RRRR'R(R)tkwargs((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyR s(                 N(R,R-R.R/R (((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyRRs-(&t __future__RRRRtnumpyRRR{R t collectionsRRRtutils.generic_utilsRtRRRxRyR/RRtobjectRR0R2R<RNRQRkRqRRRRR(((s./tmp/pip-build-isqEY4/keras/keras/callbacks.pyts>        u7CdM2koD