def setup_predict_data_feeder(X, batch_size=-1):
"""Returns an iterable for feeding into predict step.
Args:
X: numpy, pandas, Dask array or iterable.
batch_size: Size of batches to split data into.
If negative, returns one batch of full size.
Returns:
List or iterator of parts of data to predict on.
"""
if HAS_DASK:
X = extract_dask_data(X)
if HAS_PANDAS:
X = extract_pandas_data(X)
if _is_iterable(X):
return _batch_data(X, batch_size)
if len(X.shape) == 1:
X = np.reshape(X, (-1, 1))
if batch_size > 0:
n_batches = int(math.ceil(float(len(X)) / batch_size))
return [
X[i * batch_size:(i + 1) * batch_size] for i in xrange(n_batches)
]
return [X]
def _data_type_filter(X, y):
"""Filter data types into acceptable format"""
if HAS_PANDAS:
X = extract_pandas_data(X)
y = extract_pandas_labels(y)
if HAS_DASK:
X = extract_dask_data(X)
y = extract_dask_labels(y)
return X, y
def _data_type_filter(X, y):
"""Filter data types into acceptable format"""
if HAS_DASK:
X = extract_dask_data(X)
y = extract_dask_labels(y)
if HAS_PANDAS:
X = extract_pandas_data(X)
y = extract_pandas_labels(y)
return X, y
def _predict(self, X):
if not self._initialized:
raise NotFittedError()
if HAS_PANDAS:
X = extract_pandas_data(X)
if HAS_DASK:
X = extract_dask_data(X)
if len(X.shape) == 1:
X = np.reshape(X, (-1, 1))
pred = self._session.run(self._model_predictions,
feed_dict={self._inp.name: X})
return pred
def _predict(self, X):
if not self._initialized:
raise NotFittedError()
if HAS_PANDAS:
X = extract_pandas_data(X)
if HAS_DASK:
X = extract_dask_data(X)
if len(X.shape) == 1:
X = np.reshape(X, (-1, 1))
pred = self._session.run(self._model_predictions,
feed_dict={
self._inp.name: X
})
return pred
def setup_predict_data_feeder(X, batch_size=-1):
"""Returns an iterable for feeding into predict step.
Args:
X: numpy, pandas, Dask array or iterable.
batch_size: Size of batches to split data into.
If negative, returns one batch of full size.
Returns:
List or iterator of parts of data to predict on.
"""
if HAS_DASK:
X = extract_dask_data(X)
if HAS_PANDAS:
X = extract_pandas_data(X)
if _is_iterable(X):
return _batch_data(X, batch_size)
if len(X.shape) == 1:
X = np.reshape(X, (-1, 1))
return [X]