jina.executors.classifiers¶
-
class
jina.executors.classifiers.
BaseClassifier
(*args, **kwargs)[source]¶ Bases:
jina.executors.BaseExecutor
The base class of Classifier Executor. Classifier Executor allows one to perform classification and regression on given input and output the predicted hard/soft label.
This class should not be used directly. Subclasses should be used.
-
predict
(content, *args, **kwargs)[source]¶ Perform hard/soft classification on
data
, the predicted value for each sample in X is returned.The output value can be zero/one, for one-hot label; or float for soft-label or regression label. Use the corresponding driver to interpret these labels
- The size and type of output can be one of the follows,
B
isdata.shape[0]
: (B,) or (B, 1); zero/one or float
(B, L): zero/one one-hot or soft label for L-class multi-class classification
- Parameters
content (np.ndarray) – the input data to be classified, can be a ndim array. where axis=0 represents the batch size, i.e. data[0] is the first sample, data[1] is the second sample, data[n] is the n sample
args – Additional positional arguments
kwargs – Additional keyword arguments
- Return type
np.ndarray
- The size and type of output can be one of the follows,
-