def array_view(array, slicing=None, mapping=None):
"""View a slice or the entirety of an ndarray.
Args:
array: The input array, as an numpy.ndarray.
slicing: Optional slicing string, e.g., "[:, 1:3, :]".
mapping: Optional mapping string. Supported mappings:
`None` or case-insensitive `'None'`: Unmapped nested list.
`'image/png'`: Image encoding of a 2D sliced array or 3D sliced array
with 3 as the last dimension. If the sliced array is not 2D or 3D with
3 as the last dimension, a `ValueError` will be thrown.
`health-pill`: A succinct summary of the numeric values of a tensor.
See documentation in [`health_pill_calc.py`] for more details.
Returns:
1. dtype as a `str`.
2. shape of the sliced array, as a tuple of `int`s.
3. the potentially sliced values, as a nested `list`.
"""
dtype = str(array.dtype)
# String-type TensorFlow Tensors are represented as object-type arrays in
# numpy. We map the type name back to 'string' for clarity.
if dtype == 'object':
dtype = 'string'
sliced_array = (array[command_parser._parse_slices(slicing)]
if slicing else array)
if np.isscalar(sliced_array) and str(dtype) == 'string':
# When a string Tensor (for which dtype is 'object') is sliced down to only
# one element, it becomes a string, instead of an numpy array.
# We preserve the dimensionality of original array in the returned shape
# and slice.
ndims = len(array.shape)
slice_shape = []
for _ in range(ndims):
sliced_array = [sliced_array]
slice_shape.append(1)
return dtype, tuple(slice_shape), sliced_array
else:
shape = sliced_array.shape
if mapping == "image/png":
if len(sliced_array.shape) == 2:
return dtype, shape, array_to_base64_png(sliced_array)
elif len(sliced_array.shape) == 3:
raise NotImplementedError(
"image/png mapping for 3D array has not been implemented")
else:
raise ValueError("Invalid rank for image/png mapping: %d" %
len(sliced_array.shape))
elif mapping == 'health-pill':
health_pill = health_pill_calc.calc_health_pill(array)
return dtype, shape, health_pill
elif mapping is None or mapping == '' or mapping.lower() == 'none':
return dtype, shape, sliced_array.tolist()
else:
raise ValueError("Invalid mapping: %s" % mapping)
def array_view(array, slicing=None, mapping=None):
"""View a slice or the entirety of an ndarray.
Args:
array: The input array, as an numpy.ndarray.
slicing: Optional slicing string, e.g., "[:, 1:3, :]".
mapping: Optional mapping string. Supported mappings:
`None` or case-insensitive `'None'`: Unmapped nested list.
`'image/png'`: Image encoding of a 2D sliced array or 3D sliced array
with 3 as the last dimension. If the sliced array is not 2D or 3D with
3 as the last dimension, a `ValueError` will be thrown.
`health-pill`: A succinct summary of the numeric values of a tensor.
See documentation in [`health_pill_calc.py`] for more details.
Returns:
1. dtype as a `str`.
2. shape of the sliced array, as a tuple of `int`s.
3. the potentially sliced values, as a nested `list`.
"""
dtype = translate_dtype(array.dtype)
sliced_array = (array[command_parser._parse_slices(slicing)] if slicing
else array)
if np.isscalar(sliced_array) and str(dtype) == 'string':
# When a string Tensor (for which dtype is 'object') is sliced down to only
# one element, it becomes a string, instead of an numpy array.
# We preserve the dimensionality of original array in the returned shape
# and slice.
ndims = len(array.shape)
slice_shape = []
for _ in range(ndims):
sliced_array = [sliced_array]
slice_shape.append(1)
return dtype, tuple(slice_shape), sliced_array
else:
shape = sliced_array.shape
if mapping == "image/png":
if len(sliced_array.shape) == 2:
return dtype, shape, array_to_base64_png(sliced_array)
elif len(sliced_array.shape) == 3:
raise NotImplementedError(
"image/png mapping for 3D array has not been implemented")
else:
raise ValueError("Invalid rank for image/png mapping: %d" %
len(sliced_array.shape))
elif mapping == 'health-pill':
health_pill = health_pill_calc.calc_health_pill(array)
return dtype, shape, health_pill
elif mapping is None or mapping == '' or mapping.lower() == 'none':
return dtype, shape, sliced_array.tolist()
else:
raise ValueError("Invalid mapping: %s" % mapping)
def testScalar(self):
x = np.array(-1337.0)
health_pill = health_pill_calc.calc_health_pill(x)
self.assertEqual(14, len(health_pill))
self.assertEqual(1, health_pill[0]) # Is initialized.
self.assertEqual(1, health_pill[1]) # numel.
self.assertEqual(0, health_pill[2]) # NaN count.
self.assertEqual(0, health_pill[3]) # -Infinity count.
self.assertEqual(1, health_pill[4]) # Finite negative count.
self.assertEqual(0, health_pill[5]) # Zero count.
self.assertEqual(0, health_pill[6]) # Finite positive count.
self.assertEqual(0, health_pill[7]) # +Infinity count.
self.assertEqual(-1337.0, health_pill[8])
self.assertEqual(-1337.0, health_pill[9])
self.assertEqual(-1337.0, health_pill[10])
self.assertEqual(0, health_pill[11])
self.assertEqual(0, health_pill[13]) # Number of dimensions.
def testScalar(self): x = np.array(-1337.0) health_pill = health_pill_calc.calc_health_pill(x) self.assertEqual(14, len(health_pill)) self.assertEqual(1, health_pill[0]) # Is initialized. self.assertEqual(1, health_pill[1]) # numel. self.assertEqual(0, health_pill[2]) # NaN count. self.assertEqual(0, health_pill[3]) # -Infinity count. self.assertEqual(1, health_pill[4]) # Finite negative count. self.assertEqual(0, health_pill[5]) # Zero count. self.assertEqual(0, health_pill[6]) # Finite positive count. self.assertEqual(0, health_pill[7]) # +Infinity count. self.assertEqual(-1337.0, health_pill[8]) self.assertEqual(-1337.0, health_pill[9]) self.assertEqual(-1337.0, health_pill[10]) self.assertEqual(0, health_pill[11]) self.assertEqual(0, health_pill[13]) # Number of dimensions.
def testInfAndNanOnlyArray(self):
x = np.array([np.inf, -np.inf, np.nan])
health_pill = health_pill_calc.calc_health_pill(x)
self.assertEqual(15, len(health_pill))
self.assertEqual(1, health_pill[0]) # Is initialized.
self.assertEqual(3, health_pill[1]) # numel.
self.assertEqual(1, health_pill[2]) # NaN count.
self.assertEqual(1, health_pill[3]) # -Infinity count.
self.assertEqual(0, health_pill[4]) # Finite negative count.
self.assertEqual(0, health_pill[5]) # Zero count.
self.assertEqual(0, health_pill[6]) # Finite positive count.
self.assertEqual(1, health_pill[7]) # +Infinity count.
self.assertEqual(np.inf, health_pill[8])
self.assertEqual(-np.inf, health_pill[9])
self.assertTrue(np.isnan(health_pill[10]))
self.assertTrue(np.isnan(health_pill[11]))
self.assertEqual(1, health_pill[13]) # Number of dimensions.
self.assertEqual(3, health_pill[14]) # Size is (3,).
def testInfAndNanOnlyArray(self): x = np.array([np.inf, -np.inf, np.nan]) health_pill = health_pill_calc.calc_health_pill(x) self.assertEqual(15, len(health_pill)) self.assertEqual(1, health_pill[0]) # Is initialized. self.assertEqual(3, health_pill[1]) # numel. self.assertEqual(1, health_pill[2]) # NaN count. self.assertEqual(1, health_pill[3]) # -Infinity count. self.assertEqual(0, health_pill[4]) # Finite negative count. self.assertEqual(0, health_pill[5]) # Zero count. self.assertEqual(0, health_pill[6]) # Finite positive count. self.assertEqual(1, health_pill[7]) # +Infinity count. self.assertEqual(np.inf, health_pill[8]) self.assertEqual(-np.inf, health_pill[9]) self.assertTrue(np.isnan(health_pill[10])) self.assertTrue(np.isnan(health_pill[11])) self.assertEqual(1, health_pill[13]) # Number of dimensions. self.assertEqual(3, health_pill[14]) # Size is (3,).
def array_view(array, slicing=None, mapping=None):
"""View a slice or the entirety of an ndarray.
Args:
array: The input array, as an numpy.ndarray.
slicing: Optional slicing string, e.g., "[:, 1:3, :]".
mapping: Optional mapping string. Supported mappings:
`None` or case-insensitive `'None'`: Unmapped nested list.
`'image/png'`: Image encoding of a 2D sliced array or 3D sliced array
with 3 as the last dimension. If the sliced array is not 2D or 3D with
3 as the last dimension, a `ValueError` will be thrown.
`health-pill`: A succinct summary of the numeric values of a tensor.
See documentation in [`health_pill_calc.py`] for more details.
Returns:
1. dtype as a `str`.
2. shape of the sliced array, as a tuple of `int`s.
3. the potentially sliced values, as a nested `list`.
"""
dtype = str(array.dtype)
sliced_array = (array[command_parser._parse_slices(slicing)]
if slicing else array)
shape = sliced_array.shape
if mapping == "image/png":
if len(sliced_array.shape) == 2:
return dtype, shape, array_to_base64_png(sliced_array)
elif len(sliced_array.shape) == 3:
raise NotImplementedError(
"image/png mapping for 3D array has not been implemented")
else:
raise ValueError("Invalid rank for image/png mapping: %d" %
len(sliced_array.shape))
elif mapping == 'health-pill':
health_pill = health_pill_calc.calc_health_pill(array)
return dtype, shape, health_pill
elif mapping is None or mapping == '' or mapping.lower() == 'none':
return dtype, shape, sliced_array.tolist()
else:
raise ValueError("Invalid mapping: %s" % mapping)
