2VhJddlZddlZddlZddlZddlmZddlmZddl m Z ddl m Z ddl m Z edGdd e Zy) N)backend) keras_export)Callback) file_utils)io_utilszkeras.callbacks.ModelCheckpointcneZdZdZ d fd Zd dZd dZd dZdZdZ dZ d Z d Z d Z xZS)ModelCheckpointa&Callback to save the Keras model or model weights at some frequency. `ModelCheckpoint` callback is used in conjunction with training using `model.fit()` to save a model or weights (in a checkpoint file) at some interval, so the model or weights can be loaded later to continue the training from the state saved. A few options this callback provides include: - Whether to only keep the model that has achieved the "best performance" so far, or whether to save the model at the end of every epoch regardless of performance. - Definition of "best"; which quantity to monitor and whether it should be maximized or minimized. - The frequency it should save at. Currently, the callback supports saving at the end of every epoch, or after a fixed number of training batches. - Whether only weights are saved, or the whole model is saved. Example: ```python model.compile(loss=..., optimizer=..., metrics=['accuracy']) EPOCHS = 10 checkpoint_filepath = '/tmp/ckpt/checkpoint.model.keras' model_checkpoint_callback = keras.callbacks.ModelCheckpoint( filepath=checkpoint_filepath, monitor='val_accuracy', mode='max', save_best_only=True) # Model is saved at the end of every epoch, if it's the best seen so far. model.fit(epochs=EPOCHS, callbacks=[model_checkpoint_callback]) # The model (that are considered the best) can be loaded as - keras.models.load_model(checkpoint_filepath) # Alternatively, one could checkpoint just the model weights as - checkpoint_filepath = '/tmp/ckpt/checkpoint.weights.h5' model_checkpoint_callback = keras.callbacks.ModelCheckpoint( filepath=checkpoint_filepath, save_weights_only=True, monitor='val_accuracy', mode='max', save_best_only=True) # Model weights are saved at the end of every epoch, if it's the best seen # so far. model.fit(epochs=EPOCHS, callbacks=[model_checkpoint_callback]) # The model weights (that are considered the best) can be loaded as - model.load_weights(checkpoint_filepath) ``` Args: filepath: string or `PathLike`, path to save the model file. `filepath` can contain named formatting options, which will be filled the value of `epoch` and keys in `logs` (passed in `on_epoch_end`). The `filepath` name needs to end with `".weights.h5"` when `save_weights_only=True` or should end with `".keras"` or `".h5"` when checkpoint saving the whole model (default). For example: if `filepath` is `"{epoch:02d}-{val_loss:.2f}.keras"` or "{epoch:02d}-{val_loss:.2f}.weights.h5"`, then the model checkpoints will be saved with the epoch number and the validation loss in the filename. The directory of the filepath should not be reused by any other callbacks to avoid conflicts. monitor: The metric name to monitor. Typically the metrics are set by the `Model.compile` method. Note: * Prefix the name with `"val_"` to monitor validation metrics. * Use `"loss"` or `"val_loss"` to monitor the model's total loss. * If you specify metrics as strings, like `"accuracy"`, pass the same string (with or without the `"val_"` prefix). * If you pass `metrics.Metric` objects, `monitor` should be set to `metric.name` * If you're not sure about the metric names you can check the contents of the `history.history` dictionary returned by `history = model.fit()` * Multi-output models set additional prefixes on the metric names. verbose: Verbosity mode, 0 or 1. Mode 0 is silent, and mode 1 displays messages when the callback takes an action. save_best_only: if `save_best_only=True`, it only saves when the model is considered the "best" and the latest best model according to the quantity monitored will not be overwritten. If `filepath` doesn't contain formatting options like `{epoch}` then `filepath` will be overwritten by each new better model. 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 mode is set to `"max"` if the quantities monitored are `"acc"` or start with `"fmeasure"` and are set to `"min"` for the rest of the quantities. 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)`). save_freq: `"epoch"` or integer. When using `"epoch"`, the callback saves the model after each epoch. When using integer, the callback saves the model at end of this many batches. If the `Model` is compiled with `steps_per_execution=N`, then the saving criteria will be checked every Nth batch. Note that if the saving isn't aligned to epochs, the monitored metric may potentially be less reliable (it could reflect as little as 1 batch, since the metrics get reset every epoch). Defaults to `"epoch"`. initial_value_threshold: Floating point initial "best" value of the metric to be monitored. Only applies if `save_best_value=True`. Only overwrites the model weights already saved if the performance of current model is better than this value. c rt |_|_t j |_|_|_|_ d_ d_ |_ |dvrtjd|ddd}|dk(r7tj _jtj$_ n|d k(r8tj&_jtj$ _ nd jvsjj)d r8tj&_jMtj$ _ n6tj _jtj$_ jd k7r3t+jt,st/d jd|r4j j1dst/dj yt3fddDst/dj y)Nr)autominmaxzModelCheckpoint mode 'z$' is unknown, fallback to auto mode. stacklevelr r r accfmeasureepochzUnrecognized save_freq: z2. Expected save_freq are 'epoch' or integer valuesz .weights.h5zWhen using `save_weights_only=True` in `ModelCheckpoint`, the filepath provided must end in `.weights.h5` (Keras weights format). Received: filepath=c3TK|]}jj|!ywN)filepathendswith).0extselfs T/home/dcms/DCMS/lib/python3.12/site-packages/keras/src/callbacks/model_checkpoint.py z+ModelCheckpoint.__init__..s%03 &&s+s%()z.kerasz.h5zTThe filepath provided must end in `.keras` (Keras model format). Received: filepath=)super__init__monitorverboserpath_to_stringrsave_best_onlysave_weights_only save_freq_batches_seen_since_last_saving_last_batch_seenbestwarningswarnnpless monitor_opinfgreater startswith isinstanceint ValueErrorrany) rrrr r"r#moder$initial_value_threshold __class__s ` rrzModelCheckpoint.__init__s   "11(; ,!2"/0, !+ - - MM(/))  D 5= ggDOyy FF U] jjDOyy VVG  $ (?(? (K"$**99$!#DI"$''99$ "DI >>W $Z-L*4>>*:;CC  ==))-8  !% 097H! $ 0cd|j|r|j|j||yy)Nrbatchlogs)_should_save_on_batch _save_model_current_epoch)rr:r;s ron_train_batch_endz"ModelCheckpoint.on_train_batch_ends0  % %e ,   4#6#6e$  O -r7c||_yr)r>rrr;s ron_epoch_beginzModelCheckpoint.on_epoch_begins #r7cL|jdk(r|j|d|yy)Nrr9)r$r=rAs r on_epoch_endzModelCheckpoint.on_epoch_ends( >>W $   54  @ %r7c|jdk(ry||jkr|dz}n||jz }|xj|z c_||_|j|jk\rd|_yy)z?Handles batch-level saving logic, supports steps_per_execution.rFrT)r$r&r%)rr: add_batchess rr<z%ModelCheckpoint._should_save_on_batchsq >>W $ D)) )!)K$"7"77K ,, ;, %  / /4>> A34D 0r7c 8|xsi}|jrZ|j|j}|&tjd|jddyt |t jstj|r2t|jdkDrtjd|dy|j||jrT|jdkDr=tj d |d zd |jd |jd d|d d| ||_ y|jdkDr6tj d |d zd |jd|jd y|jdkDrtj d |d zd|y)aDetermines whether the model should be saved. The model should be saved in the following cases: - self.save_best_only is False - self.save_best_only is True and `monitor` is a numpy array or backend tensor (falls back to `save_best_only=False`) - self.save_best_only is True and `self.monitor_op(current, self.best)` evaluates to True. Args: epoch: the epoch this iteration is in. batch: the batch this iteration is in. `None` if the `save_freq` is set to `"epoch"`. logs: the `logs` dict passed in to `on_batch_end` or `on_epoch_end`. filepath: the path where the model would be saved zCan save best model only with z available.rrTrzECan save best model only when `monitor` is a scalar value. Received: z). Falling back to `save_best_only=False`.z Epoch rFz: z improved from z.5fz to z, saving model to z did not improve from Fz: saving model to )r"getrr(r)r0r*ndarrayr is_tensorlenshaper,r'r r print_msg)rrr:r;rcurrents r_should_save_modelz"ModelCheckpoint._should_save_models&zr   hht||,G 4T\\N+N 7BJJ/73D3DW3Mgmm$q( 118 :>> ??7DII6||a' **&uqykDLL>B$$(IIc?$wsmD//7j: !(DI||a' **&uqyk#||n-$$(IIc?4 !||a""uqyk);H:Fr7c8|j|||} |j||||rtjj |}|r*t j |st j||jr|jj|dy|jj|dyy#t$rtd|t$r=}dt|jdj!vrtd||d}~wwxYw)aSaves the model. Args: epoch: the epoch this iteration is in. batch: the batch this iteration is in. `None` if the `save_freq` is set to `"epoch"`. logs: the `logs` dict passed in to `on_batch_end` or `on_epoch_end`. T) overwritezePlease specify a non-directory filepath for ModelCheckpoint. Filepath used is an existing directory: zis a directoryrzfPlease specify a non-directory filepath for ModelCheckpoint. Filepath used is an existing directory: fN)_get_file_pathrPospathdirnamerexistsmakedirsr#model save_weightssaveIsADirectoryErrorIOErrorstrargslower)rrr:r;rrVes rr=zModelCheckpoint._save_model"s&&ueT: &&ueT8D''//(3:#4#4W#=''0))JJ++H+EJJOOHO=E! &Z)    3qvvay>#7#7#99##+*. G sBB=B==D8DDc |d|vr#|jjdd|dzi|}|S|jjd|dz|dzd|} |S#t$r }td|jd|d}~wwxYw) z%Returns the file path for checkpoint.Nr:rrF)rr:z*Failed to format this callback filepath: "z ". Reason: )rformatKeyError)rrr:r; file_pathras rrSzModelCheckpoint._get_file_pathJs  }40DMM00IuqyIDI 1DMM00!)5198<   ` operator, lexicographically using the numeric equivalents). This provides a tie-breaker when multiple files are most recent. Note that a larger `filepath` can sometimes indicate a later time of modification (for instance, when epoch/batch is used as formatting option), but not necessarily (when accuracy or loss is used). The tie-breaker is put in the logic as best effort to return the most recent, and to avoid nondeterministic result. Modified time of a file is obtained with `os.path.getmtime()`. This utility function is best demonstrated via an example: ```python file_pattern = 'batch{batch:02d}epoch{epoch:02d}.keras' test_dir = self.get_temp_dir() path_pattern = os.path.join(test_dir, file_pattern) file_paths = [ os.path.join(test_dir, file_name) for file_name in ['batch03epoch02.keras', 'batch02epoch02.keras', 'batch01epoch01.keras'] ] for file_path in file_paths: # Write something to each of the files ... self.assertEqual( _get_most_recently_modified_file_matching_pattern(path_pattern), file_paths[-1]) ``` Args: pattern: The file pattern that may optionally contain python placeholder such as `{epoch:02d}`. Returns: The most recently modified file's full filepath matching `pattern`. If `pattern` does not contain any placeholder, this returns the filepath that exactly matches `pattern`. Returns `None` if no match is found. ^z{.*}z.*$rNrF) rTrUrVbasenameresubrrWlistdirmatchjoingetmtime) rpatterndir_name base_namebase_name_regexlatest_mod_timefile_path_with_latest_mod_timen_file_with_latest_mod_time file_path_with_largest_file_name file_namerfmod_times r1_get_most_recently_modified_file_matching_patternzAModelCheckpoint._get_most_recently_modified_file_matching_patterncsX77??7+GG$$W- wy AACG)-&&'#+/(   X &ZZ1 9 88OY7 " Xy AI!ww// :H8@$'GG;D8/1*29B6783!_44q83+ 9. '! +1 14 3r7)val_lossrFFr rNr)__name__ __module__ __qualname____doc__rr?rBrDr<rPr=rSrhr} __classcell__)r6s@rr r sZnf  $FPP$A =~&P*+S4r7r )rTrmr(numpyr* keras.srcrkeras.src.api_exportrkeras.src.callbacks.callbackrkeras.src.utilsrrr rcr7rrsD -1&$/0g4hg41g4r7