o
M eyx @ sx d Z ddlZddlZddlm mZ ddlm Z ddlm
Z ddl m
Z
ddlmZ ddlmZ ddlmZ dd lmZ dd
lmZ dd lmZ zdd lmZ W n e y\ dZY nw dddd
d
ddddddddd
ej dddfddZ dd Z dd Z dd Z!dd Z"d'dd Z#d'd d Z$d d Z%d!d" Z&ej'e ej d#Z(ej'e ej)dd$Z*ej'e ej+dd$Z,G d%d& d&ej-Z.dS )(z>Part of the Keras training engine related to plain array data. N)backend) callbacks) distributed_training_utils_v1)training_utils_v1)io_utils) make_batches) slice_arrays)ModeKeys)
tf_logging)issparse TFstepsc 7 K sD d|v r | d}
|rtd| d}d}t|tjjjjtjjf}|r;|
du r5d}tj | ||
||d}
t
|| j }| j rQt j
| j |tjkrIdndd }| |pV|
du}|du}|p^|}|rf|rf|}nt| ||||}|swt|||
} n|
} t| || t| |} d} t|tjjjjtjjfr|du rtj | |||d
d}t
|| j } t| | | |
tj}|} n|rtj|d jd pd} |tjkr|rt| | | |rdnd } tj|| ||||
| | ||d
}t dur|sg }!t | |}"t t ||"D ]\}#\}$}%t |$rt !|%s|!"|# q|tj#kr%tj$||
r dn| |
d
}&n
tj%||
r-dn| |
d
}&| j&r| }.| dkrCt?|.|}.n | rLt/j@A|. tB| |}/t |/D ]\}0\}1}2|.|1|2 }3tC|/dkrk|}4n.z!|rt|d tDrtE|dd |3|d g }4ntE||3}4W n ty tdw t dur|!D ] }#|4|# F |4|#< q|0tC|3d}+|1|d|0|+ | |4}-t|-t9s|-g}-|0dkr|&;|- |&<|-|1|2 |=| |+|-|}+|1|d|0|+ |j(j)r nqU|&G |&jH}5|=| |(|5|}(tC|5dkr |5d }5|r\tI||'r\|j(j)s\| j&r&t J| tj tK| || |
|||dtjd| dud
d }6t|6t9sB|6g}6|j=| |(|6|dd}(| r\|'|d k r\tL| | j |tjkrh|M|'|( |rx|'|d k rxtL|| j qPd| _N|O| | j r| j&rt J| | |Pddd |tjkr| jQS |5S )a8
Loop function for arrays of data with modes TRAIN/TEST/PREDICT.
Args:
model: Keras Model instance.
inputs: Either a list or dictionary of arrays, or a dataset instance.
targets: List/dictionary of input arrays.
sample_weights: Optional list of sample weight arrays.
batch_size: Integer batch size or None if unknown.
epochs: Number of times to iterate over the data
verbose: 0, 1, or 2. Verbosity mode.
0 = silent, 1 = progress bar, 2 = one line per epoch.
Note that the progress bar is not particularly useful when
logged to a file, so verbose=2 is recommended when not running
interactively (eg, in a production environment).
callbacks: List of callbacks to be called during training
val_inputs: Either a list or dictionary of arrays, or a dataset
instance.
val_targets: List/dictionary of target arrays.
val_sample_weights: Optional list of sample weight arrays.
shuffle: Whether to shuffle the data at the beginning of each epoch
concatenation of list the display names of the outputs of `f` and the
list of display names of the outputs of `f_val`.
initial_epoch: Epoch at which to start training (useful for resuming a
previous training run)
steps_per_epoch: Total number of steps (batches of samples) before
declaring one epoch finished and starting the next epoch. Ignored with
the default value of `None`.
validation_steps: Number of steps to run validation for (only if doing
validation from data tensors). Ignored with the default value of
`None`.
validation_freq: Only relevant if validation data is provided. Integer
or `collections.abc.Container` instance (e.g. list, tuple, etc.). If
an integer, specifies how many training epochs to run before a new
validation run is performed, e.g. `validation_freq=2` runs validation
every 2 epochs. If a Container, specifies the epochs on which to run
validation, e.g. `validation_freq=[1, 2, 10]` runs validation at the
end of the 1st, 2nd, and 10th epochs.
mode: One of ModeKeys.TRAIN/ModeKeys.TEST/ModeKeys.PREDICT.
validation_in_fit: if true, then this method is invoked from within
training iteration (for validation). In the case where `val_inputs` is
a dataset, this flag indicates that its iterator and feed values are
already created so should properly reuse resources.
prepared_feed_values_from_dataset: if True, `inputs` is a list of feed
tensors returned from `_prepare_feed_values` call on the validation
dataset, so do not call it again on `inputs`. Should only be used for
inline validation (i.e., only if `validation_in_fit` is also True).
steps_name: The string name of the steps argument, either `steps`,
`validation_steps`, or `steps_per_epoch`. Only used for error message
formatting.
**kwargs: Additional arguments for backwards compatibility.
Returns:
- In TRAIN mode: `History` object.
- In TEST mode: Evaluation metrics.
- In PREDICT mode: Outputs of the Model called on inputs.
Raises:
ValueError: in case of invalid arguments.
r
zUnknown arguments: FNT)epochs
steps_namer r )ZstrategyZlearning_phasevalidation_stepssamples)
do_validation
batch_sizer steps_per_epochr
count_modeverbosemode)Z num_samplesr
)batchsizebeginzYour dataset ran out of data; interrupting training. Make sure that your dataset can generate at least `%s * epochs` batches (in this case, %d batches). You may need to use the repeat() function when building your dataset.zYour dataset iterator ran out of data; interrupting training. Make sure that your iterator can generate at least `%s * epochs` batches (in this case, %d batches). You may need to use the repeat() function when building your dataset.endr zPTypeError while preparing batch. If using HDF5 input data, pass shuffle="batch".)
targetssample_weightsr r r r r validation_in_fit!prepared_feed_values_from_datasetr Zval_)prefix)Rpop TypeError
isinstancetfcompatv1dataDatasetr Zinfer_steps_for_dataset
_get_iterator_distribution_strategyr Zdistributed_scoper TRAIN __enter___prepare_feed_values_get_num_samples_or_steps_update_sample_weight_mode_make_execution_functionTESTnestflattenshape_print_train_infocbksZconfigure_callbacksr _get_model_feed enumeratezipr Z is_sparseappendPREDICTZOutputsAggregatorZMetricsAggregatorZ_compile_distributionZ"_copy_weights_to_distributed_modelmodelZ
stop_trainingZ_call_begin_hookZ#_maybe_load_initial_epoch_from_ckptrangeZ
reset_metricsZon_epoch_beginnpinfZ_call_batch_hookcallableZis_distributing_by_cloningerrorsZOutOfRangeErrorloggingwarningr
listZ _per_replica_aggregate_batchcreateZ aggregateZ make_logsZarangeZ
batch_shufflerandomshuffler lenintr ZtoarrayfinalizeresultsZshould_run_validationZ _copy_weights_to_original_modelmodel_iteration_reinitialize_iteratorZ on_epoch_endZ_successful_loop_finishZ_call_end_hook__exit__history)7r= inputsr r r r r r
val_inputs val_targetsval_sample_weightsrH
initial_epochr r validation_freqr r r r kwargsZ reset_dataset_after_each_epochZinput_iterator
is_datasetscopeZ use_stepsr insnum_samples_or_stepsfZ val_iteratorval_samples_or_stepsr Z indices_for_conversion_to_densefeediZ
input_dataZ feed_tensorZ
aggregatorepochZ
epoch_logsZ target_stepsstepZ
batch_logsZ
actual_inputsZ
batch_outsZ index_arrayZbatchesZ batch_indexZ batch_startZ batch_endZ batch_idsZ ins_batchrL Z val_results rb OD:\Projects\ConvertPro\env\Lib\site-packages\keras/engine/training_arrays_v1.pyrM ) s S
#
M
rM c C s* |t jkr
| j}|S | j| j | j }|S N)r r< _feed_inputs
_feed_targetsZ_feed_sample_weights)r= r r^ rb rb rc r8 s
r8 c C sB |rdnd}d| d| }|r|d| d| 7 }t | d S )Nr
r z Train on z, validate on )r Z print_msg)r[ r] rX incrementmsgrb rb rc r6 s
r6 c C s |r|S t | ||dS )zEReturns total number of samples when training in batch mode or steps.r )r Zcheck_num_samples)rZ r r rb rb rc r/ s
r/ c s j r*t tjjjjtjjfrt j fdd}t r'|S | S t tjjjjtjjtjjjj
frFj dd\ t
tpQg tpWg }tjkrptt tsp|dg7 }|S )a Prepare feed values to the model execution function.
Args:
model: Model to prepare feed values for.
inputs: List or dict of model inputs.
targets: Optional list of model targets.
sample_weights: Optional list of sample weight arrays.
mode: One of ModeKeys.TRAIN/ModeKeys.TEST/ModeKeys.PREDICT.
Returns:
Feed values for the model in the given mode.
c s t S rd )r r. rb rQ r r= r r rb rc get_distributed_inputs- s
z4_prepare_feed_values..get_distributed_inputsT)Z extract_tensors_from_dataset)r+ r$ r% r& r' r( r) r get_iteratorZexecuting_eagerlyIterator_standardize_user_datar Z ModelInputsas_listrE r r, r symbolic_learning_phaserJ )r= rQ r r r rk rZ rb rj rc r. s8
r. c C s |rt | |S t| S rd )r rl r )rQ distribution_strategyrb rb rc r* U s
r* c C s" |r
t | | d S t| d S rd )r Zinitialize_iteratorr )iteratorrq rb rb rc rN ] s
rN c C s | j r t| |S | |S )z2Makes function to run one step of model execution.)r+ r r1 )r= r rb rb rc r1 f s
r1 c C s |t jkrdS d}t|s6|t| jt| j d }|t jko'tt t
}|r0|dd }| j |d | j rBt
| || dS dS )z0Updates the sample_weight_mode of a given model.Nr )r )r r< rA rI re rf r, r$ r rp rJ Z_update_sample_weight_modesr+ r )r= r rQ r Zhas_learning_phase_plrb rb rc r0 o s$
r0 )r )r rH c @ sb e Zd ZdZ d
ddZ dd d
Z dd d ZdS )ArrayLikeTrainingLoopa= TrainingLoop that handle inputs like array.
This is the default handler for most of the input data types, includes
symbolic tensors or Numpy array-like, Datasets and iterators in graph mode
(since they generate symbolic tensors). This Function is used to handle
model with `run_eagerly` = False.
Nr Tr c K s | |||}|j||| | |dd|||
d
\}}}| r&|| ||\}}}n(|rCd| k r2dk rCn nt||||\}}}}}}n |rItdd\}}}t|fi d|d |d
|d |d |d
|d|d|d|d|d|
d|
d|d|d|ddS )NTr )
sample_weight class_weightr check_stepsr r
validation_splitrH rt g ?zH`validation_steps` should not be specified if `validation_data` is None.)NNNrQ r r r r r r rR rS rT rH rU r rV r ) _validate_or_infer_batch_sizern Z_prepare_validation_datar Z"split_training_and_validation_data
ValueErrorfit_loop)selfr= xyr r r r rx Zvalidation_datarH rv ru rU r r rV rW r Zval_xZval_yrT rb rb rc fit s
zArrayLikeTrainingLoop.fitc
K sD | |||}|j||||dd|d\}}}
t||||
||||dS )NTr
)ru r rw r r
)rQ r r r r r
r )ry rn test_loop) r| r= r} r~ r r ru r
r rW r rb rb rc evaluate s( z ArrayLikeTrainingLoop.evaluatec K s: | |||}|j|dd|d\}}}t||||||dS )NTr
)rw r r
)r r r
r )ry rn predict_loop) r| r= r} r r r
r rW _rb rb rc predict s
z ArrayLikeTrainingLoop.predict)NNNr r Nrt NTNNr NNr )NNNr NNN)Nr NN)__name__
__module__ __qualname____doc__r r r rb rb rb rc rs s>
S
%rs rd )/r functoolsnumpyr? Ztensorflow.compat.v2r& v2r% Zkerasr r r7 Zkeras.distributer Z keras.enginer Z keras.utilsr Zkeras.utils.generic_utilsr r Zkeras.utils.mode_keysr Ztensorflow.python.platformr
rC Z scipy.sparser ImportErrorr, rM r8 r6 r/ r. r* rN r1 r0 partialr{ r2 r r< r Z TrainingLooprs rb rb rb rc sp
W
;
!