def _feed_dict_fn():
# TODO: option for with/without replacement (dev version of dask)
sample = self.df.random_split([self.sample_fraction, 1-self.sample_fraction],
random_state=self.random_state)
inp = extract_pandas_matrix(sample[0][self.X_columns].compute()).tolist()
out = extract_pandas_matrix(sample[0][self.y_columns].compute()).tolist()
return {input_placeholder.name: inp, output_placeholder.name: out}
def _feed_dict_fn():
# TODO: option for with/without replacement (dev version of dask)
sample = self.df.random_split([self.sample_fraction, 1-self.sample_fraction],
random_state=self.random_state)
inp = extract_pandas_matrix(sample[0][self.X_columns].compute()).tolist()
out = extract_pandas_matrix(sample[0][self.y_columns].compute())
# convert to correct dtype
inp = np.array(inp, dtype=self.input_dtype)
# one-hot encode out for each class for cross entropy loss
if HAS_PANDAS:
import pandas as pd
if not isinstance(out, pd.Series):
out = out.flatten()
out_max = self.y.max().compute().values[0]
encoded_out = np.zeros((out.size, out_max+1), dtype=self.output_dtype)
encoded_out[np.arange(out.size), out] = 1
return {input_placeholder.name: inp, output_placeholder.name: encoded_out}
def _feed_dict_fn():
# TODO: option for with/without replacement (dev version of dask)
sample = self.df.random_split([self.sample_fraction, 1-self.sample_fraction],
random_state=self.random_state)
inp = extract_pandas_matrix(sample[0][self.X_columns].compute()).tolist()
out = extract_pandas_matrix(sample[0][self.y_columns].compute())
# convert to correct dtype
inp = np.array(inp, dtype=self.input_dtype)
# one-hot encode out for each class for cross entropy loss
if HAS_PANDAS:
import pandas as pd
if not isinstance(out, pd.Series):
out = out.flatten()
out_max = self.y.max().compute().values[0]
encoded_out = np.zeros((out.size, out_max+1), dtype=self.output_dtype)
encoded_out[np.arange(out.size), out] = 1
return {input_placeholder.name: inp, output_placeholder.name: encoded_out}
def setup_processor_data_feeder(X):
"""Sets up processor iterable.
Args:
X: numpy, pandas or iterable.
Returns:
Iterable of data to process.
"""
if HAS_PANDAS:
X = extract_pandas_matrix(X)
return X
def setup_processor_data_feeder(X):
"""Sets up processor iterable.
Args:
X: numpy, pandas or iterable.
Returns:
Iterable of data to process.
"""
if HAS_PANDAS:
X = extract_pandas_matrix(X)
return X
