o 2hX@s@ddlZddlmZddlmZddlmZGdddeZdS)N)ops)Callback) compile_utilscs:eZdZdZ    d fdd Zdd Zd d ZZS) MonitorCallbackaBase class for callbacks that monitor a quantity and evaluates improvements. This class provides common functionality for callbacks that monitor a metric during training to determine whether a condition has been met, such as improvement over time. It encapsulates logic for selecting the comparison operation based on a `monitor` value and `mode`, and computing whether a new value is an improvement. It is intended to be subclassed by other callbacks like `ModelCheckpoint`, `EarlyStopping`, or `ReduceLROnPlateau`, and is not meant to be used directly. Arguments: monitor: Quantity to be monitored. Defaults to `"val_loss"`. mode: One of `{"auto", "min", "max"}`. In `min` mode, training will aim to minimize the monitored quantity; in `'max'` mode it will aim to maximize it.; in `"auto"` mode, the direction is automatically inferred from the name of the monitored quantity. Defaults to `"auto"`. baseline: Floating point initial "best" value of the metric to be monitored. If `None` (default), the first monitored value will be used. 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. Defaults to `0`. Raises: ValueError: If `mode='auto'` is selected and the direction of the metric cannot be inferred. val_lossautoNrcsZt|dvrtj|jjd|dddd}||_||_||_t ||_ d|_ dS)N)rminmaxz mode 'z$' is unknown, fallback to auto mode.) stacklevelr) super__init__warningswarn __class____name__monitormodebestabs min_delta monitor_op)selfrrbaselinerr_/var/www/html/chatgem/venv/lib/python3.10/site-packages/keras/src/callbacks/monitor_callback.pyr )s   zMonitorCallback.__init__cCs|jdkr tj|_ne|jdkrtj|_n[|jd}|dkr"tj|_t|jdr]g}|jj D]}t |t j t j fr?||j q.|D]}|j|kr\t|dr\|jdkrXtj|_qBtj|_qB|jdurot|jjd|jd |jtjkr~|jd 9_dSdS) Nrr val_lossmetrics _directionupz callback received monitor=z, but Keras isn't able to automatically determine whether that metric should be maximized or minimized. Pass `mode='max'` in order to monitor based on the highest metric value, or pass `mode='min'` in order to use the lowest value.)rrlessrgreaterr removeprefixhasattrmodelr isinstancerCompileMetrics MetricsListextendnamer ValueErrorrrr)r metric_name all_metricsmrrr_set_monitor_op>sD               zMonitorCallback._set_monitor_opcCs|durdS|||j|S)NT)rr)r monitor_valuereference_valuerrr_is_improvementeszMonitorCallback._is_improvement)rrNr)r __module__ __qualname____doc__r r1r4 __classcell__rrrrrs"'r)r keras.srcrkeras.src.callbacks.callbackrkeras.src.trainersrrrrrrs