def WrapConstants(constants, dtype='float32'):
if not isinstance(constants, Tensor):
if not isinstance(constants, np.ndarray):
constants = np.array(constants, dtype=dtype)
tensor = Tensor()
tensor.set_value(constants)
tensor.shape = constants.shape
constants = tensor
return constants
def constant(value, dtype=None, shape=None, name=None):
if dtype == None: dtype = dtypes.float32
if isinstance(value, np.ndarray): feed = value.astype(dtype)
elif isinstance(value, list): feed = np.array(value, dtype)
else: feed = np.array([value], dtype)
if shape is not None:
if feed.size == 1:
c = feed[0]
feed = np.zeros(shape, dtype)
feed.fill(c)
else: feed = feed.reshape(shape)
tensor = Tensor(name)
tensor.shape = list(feed.shape)
ws.FeedTensor(tensor, feed)
return tensor
def constant(value, dtype=None, shape=None, name=None, verify_shape=False):
# determine the data type
if dtype == None: dtype = dtypes.float32
if isinstance(value, np.ndarray):
feed = value.astype(dtype.as_numpy_dtype)
elif isinstance(value, list):
feed = np.array(value, dtype.as_numpy_dtype)
else:
feed = np.array([value], dtype.as_numpy_dtype)
# determine the shape
if shape is not None:
if feed.size == 1:
# case 1: broadcast with scalar value
c = feed[0]
feed = np.zeros(shape, dtype.as_numpy_dtype)
feed.fill(c)
else:
# case 2: reshape directly
if verify_shape:
if shape is not None:
if len(shape) != len(value.shape):
raise RuntimeError(
'The constant was limited to {} dimensions, \
while feed a value with {} dimensions.'
.format(len(shape), len(value.shape)))
for i in xrange(len(shape)):
if shape[i] is None: continue
if shape[i] != value.shape[i]:
raise RuntimeError(
'The shape of constant was limited as (' +
','.join([str(dim) for dim in shape]) + '), ' +
'while feed a value with (' +
','.join([str(dim)
for dim in value.shape]) + ').')
feed = feed.reshape(shape)
# feed to VM
tensor = Tensor(name)
tensor.shape = list(feed.shape)
ws.FeedTensor(tensor, feed)
return tensor
def constant(value, dtype=None, shape=None, name=None, verify_shape=False):
# determine the data type
if dtype == None: dtype = dtypes.float32
if isinstance(value, np.ndarray):
feed = value.astype(dtype.as_numpy_dtype)
elif isinstance(value, list):
feed = np.array(value, dtype.as_numpy_dtype)
else:
feed = np.array([value], dtype.as_numpy_dtype)
# determine the shape
if shape is not None:
if feed.size == 1:
# case 1: broadcast with scalar value
c = feed[0]
feed = np.zeros(shape, dtype.as_numpy_dtype)
feed.fill(c)
else:
# case 2: reshape directly
if verify_shape:
if shape is not None:
if len(shape) != len(value.shape):
raise RuntimeError('The constant was limited to {} dimensions, \
while feed a value with {} dimensions.'.
format(len(shape), len(value.shape)))
for i in xrange(len(shape)):
if shape[i] is None: continue
if shape[i] != value.shape[i]:
raise RuntimeError('The shape of constant was limited as (' +
','.join([str(dim) for dim in shape]) + '), ' +
'while feed a value with (' + ','.join([str(dim) for dim in value.shape]) + ').')
feed = feed.reshape(shape)
# feed to VM
tensor = Tensor(name)
tensor.shape = list(feed.shape)
ws.FeedTensor(tensor, feed)
return tensor
