def as_recarray(info_blob):
"""
Save network output as recarray to h5. Intended for when network
outputs are 2D, i.e. (batchsize, X).
Output from network:
Dict with arrays, shapes (batchsize, x_i).
E.g. {"foo": ndarray, "bar": ndarray}
dtypes that will get saved to h5:
(foo_1, foo_2, ..., bar_1, bar_2, ... )
"""
datasets = dict()
datasets["pred"] = misc.dict_to_recarray(info_blob.get("y_pred"))
ys = info_blob.get("ys")
if ys is not None:
datasets["true"] = misc.dict_to_recarray(ys)
y_values = info_blob.get("y_values")
if y_values is not None:
datasets['y_values'] = y_values # is already a structured array
return datasets
def test_dict_to_recarray_weird_shape(self):
inp = {"aa": np.ones((5, 3, 2)), "bb": np.ones((5, 1))}
output = misc.dict_to_recarray(inp)
self.assertTrue(output.shape == (5, ))
self.assertTupleEqual(
output.dtype.names,
('aa_1', 'aa_2', 'aa_3', 'aa_4', 'aa_5', 'aa_6', 'bb_1'))
def test_dict_to_recarray_dtype(self):
inp = {"aa": np.ones((5, 3), dtype="int16"), "bb": np.ones((5, 1), dtype="float32")*4}
output = misc.dict_to_recarray(inp)
dtypes = {
"aa_1": np.dtype("int16"),
"aa_2": np.dtype("int16"),
"aa_3": np.dtype("int16"),
"bb_1": np.dtype("float32"),
}
self.assertTupleEqual(output.dtype.names, ('aa_1', 'aa_2', 'aa_3', 'bb_1'))
for dtype_name, dtype in dtypes.items():
np.testing.assert_equal(output[dtype_name].dtype, dtype)
np.testing.assert_equal(output["aa_1"], np.ones(5, dtype="int16"))
np.testing.assert_equal(output["bb_1"], np.ones(5, dtype="float32")*4)
def test_dict_to_recarray_shape(self):
inp = {"aa": np.ones((5, 3)), "bb": np.ones((5, 1))}
output = misc.dict_to_recarray(inp)
self.assertTrue(output.shape == (5, ))
def test_dict_to_recarray_wrong_dim(self):
inp = {"aa": np.ones((4, 3)), "bb": np.ones((5, 1))}
with self.assertRaises(ValueError):
misc.dict_to_recarray(inp)