def from_csv_with_feature_spec(cls,
filepatterns,
feature_spec,
has_header=True,
column_names=None,
num_threads=1,
enqueue_size=None,
batch_size=32,
queue_capacity=None,
min_after_dequeue=None,
shuffle=True,
seed=None):
"""Create a `DataFrame` from CSV files, given a feature_spec.
If `has_header` is false, then `column_names` must be specified. If
`has_header` is true and `column_names` are specified, then `column_names`
overrides the names in the header.
Args:
filepatterns: a list of file patterns that resolve to CSV files.
feature_spec: a dict mapping column names to `FixedLenFeature` or
`VarLenFeature`.
has_header: whether or not the CSV files have headers.
column_names: a list of names for the columns in the CSV files.
num_threads: the number of readers that will work in parallel.
enqueue_size: block size for each read operation.
batch_size: desired batch size.
queue_capacity: capacity of the queue that will store parsed lines.
min_after_dequeue: minimum number of elements that can be left by a
dequeue operation. Only used if `shuffle` is true.
shuffle: whether records should be shuffled. Defaults to true.
seed: passed to random shuffle operations. Only used if `shuffle` is true.
Returns:
A `DataFrame` that has columns corresponding to `features` and is filled
with examples from `filepatterns`.
Raises:
ValueError: no files match `filepatterns`.
ValueError: `features` contains the reserved name 'index'.
"""
def get_default_values(column_names):
return [_get_default_value(feature_spec[name]) for name in column_names]
dataframe = cls._from_csv_base(filepatterns, get_default_values, has_header,
column_names, num_threads,
enqueue_size, batch_size, queue_capacity,
min_after_dequeue, shuffle, seed)
# replace the dense columns with sparse ones in place in the dataframe
for name in dataframe.columns():
if name != "index" and isinstance(feature_spec[name],
parsing_ops.VarLenFeature):
strip_value = _get_default_value(feature_spec[name])
(dataframe[name],) = sparsify.Sparsify(strip_value)(dataframe[name])
return dataframe
def _test_sparsify_densify(self, x, default_value):
"""Test roundtrip via Sparsify and Densify."""
numpy_source = in_memory_source.NumpySource(x, batch_size=len(x))()
(sparse_series, ) = sparsify.Sparsify(default_value)(numpy_source[1])
(dense_series, ) = densify.Densify(default_value)(sparse_series)
cache = {}
sparse_tensor = sparse_series.build(cache)
dense_tensor = dense_series.build(cache)
with self.test_session() as sess:
coord = tf.train.Coordinator()
threads = tf.train.start_queue_runners(sess=sess, coord=coord)
sparse_val, dense_val = sess.run([sparse_tensor, dense_tensor])
coord.request_stop()
coord.join(threads)
if x.dtype.kind not in ["S", "U"] and np.isnan(default_value):
x_values = x[~np.isnan(x)]
x_indexes = np.arange(len(x))[~np.isnan(x)].T.reshape(-1, 1)
else:
x_values = x[x != default_value]
x_indexes = np.arange(len(x))[x != default_value].T.reshape(-1, 1)
if x.dtype.kind in ["S", "U"]:
# Python 2/3 compatibility
# TensorFlow always returns bytes, so we just convert the unicode
# expectations to bytes also before comparing.
expected_x = [item.encode("utf-8") for item in x]
expected_x_values = [item.encode("utf-8") for item in x_values]
else:
expected_x = x
expected_x_values = x_values
np.testing.assert_array_equal(len(x), sparse_val.shape[0])
np.testing.assert_array_equal(expected_x_values, sparse_val.values)
np.testing.assert_array_equal(x_indexes, sparse_val.indices)
np.testing.assert_array_equal(expected_x, dense_val)