def get_seed(op_seed, kernel_name):
"""
Get the graph-level seed.
Graph-level seed is used as a global variable, that can be used in different ops in case op-level seed is not set.
If op-level seed is 0, use graph-level seed; if graph-level seed is also 0, the system would generate a
random seed.
Note:
For each seed, either op-seed or graph-seed, a random sequence will be generated relating to this seed.
So, the state of the seed regarding to this op should be recorded.
A simple illustration should be:
If a random op is called twice within one program, the two results should be different:
print(C.uniform((1, 4), seed=1)) # generates 'A1'
print(C.uniform((1, 4), seed=1)) # generates 'A2'
If the same program runs again, it repeat the results:
print(C.uniform((1, 4), seed=1)) # generates 'A1'
print(C.uniform((1, 4), seed=1)) # generates 'A2'
Returns:
Interger. The current graph-level seed.
Examples:
>>> C.get_seed(seed, 'normal')
"""
global_seed = get_global_seed()
if global_seed is None:
global_seed = 0
if op_seed is None:
temp_seed = _get_op_seed(0, kernel_name)
else:
Validator.check_non_negative_int(op_seed, "seed", kernel_name)
temp_seed = _get_op_seed(op_seed, kernel_name)
seeds = _truncate_seed(global_seed), _truncate_seed(temp_seed)
_update_seeds(op_seed, kernel_name)
return seeds
def __init__(self,
seed,
dtype,
name,
param):
"""
Constructor of distribution class.
"""
super(Distribution, self).__init__()
if seed is None:
seed = 0
validator.check_value_type('name', name, [str], type(self).__name__)
validator.check_non_negative_int(seed, 'seed', name)
self._name = name
self._seed = seed
self._dtype = cast_type_for_device(dtype)
self._parameters = {}
# parsing parameters
for k in param.keys():
if not(k == 'self' or k.startswith('_')):
self._parameters[k] = param[k]
# some attributes
if 'distribution' in self.parameters.keys():
self.parameter_type = self.parameters['distribution'].parameter_type
else:
self.parameter_type = set_param_type(self.parameters['param_dict'], dtype)
self._broadcast_shape = self._calc_broadcast_shape()
self._is_scalar_batch = self._check_is_scalar_batch()
# set the function to call according to the derived class's attributes
self._set_prob()
self._set_log_prob()
self._set_sd()
self._set_var()
self._set_cdf()
self._set_survival()
self._set_log_cdf()
self._set_log_survival()
self._set_cross_entropy()
self.context_mode = context.get_context('mode')
self.device_target = context.get_context('device_target')
self.checktuple = CheckTuple()
self.checktensor = CheckTensor()
self.broadcast = broadcast_to
# ops needed for the base class
self.cast_base = P.Cast()
self.dtype_base = P.DType()
self.exp_base = exp_generic
self.fill_base = P.Fill()
self.log_base = log_generic
self.sametypeshape_base = P.SameTypeShape()
self.sq_base = P.Square()
self.sqrt_base = P.Sqrt()
self.shape_base = P.Shape()
def __init__(self,
in_channels,
out_channels,
kernel_size,
stride=1,
pad_mode='same',
padding=0,
dilation=1,
group=1,
has_bias=False,
weight_init='normal',
bias_init='zeros'):
Validator.check_value_type("kernel_size", kernel_size, [int],
self.cls_name)
Validator.check_value_type("stride", stride, [int], self.cls_name)
Validator.check_value_type("padding", padding, [int], self.cls_name)
Validator.check_value_type("dilation", dilation, [int], self.cls_name)
Validator.check_int(kernel_size, 1, Rel.GE, 'kernel_size',
self.cls_name)
Validator.check_int(stride, 1, Rel.GE, 'stride', self.cls_name)
Validator.check_non_negative_int(padding, 'padding', self.cls_name)
Validator.check_int(dilation, 1, Rel.GE, 'dilation', self.cls_name)
kernel_size = (1, kernel_size)
stride = (1, stride)
dilation = (1, dilation)
get_shape = P.Shape()
get_dtype = P.DType()
if isinstance(weight_init, Tensor):
weight_init_shape = get_shape(weight_init)
Validator.check_equal_int(len(weight_init_shape), 3,
'weight_init_shape', self.cls_name)
weight_init_dtype = get_dtype(weight_init)
weight_init_value = weight_init.asnumpy()
weight_init_value = np.expand_dims(weight_init_value, 2)
weight_init = Tensor(weight_init_value, weight_init_dtype)
super(Conv1d, self).__init__(in_channels, out_channels, kernel_size,
stride, pad_mode, padding, dilation,
group, has_bias, weight_init, bias_init)
self.padding = (0, 0, padding, padding)
self.conv2d = P.Conv2D(out_channel=self.out_channels,
kernel_size=self.kernel_size,
mode=1,
pad_mode=self.pad_mode,
pad=self.padding,
stride=self.stride,
dilation=self.dilation,
group=self.group)
self.bias_add = P.BiasAdd()
if pad_mode not in ('valid', 'same', 'pad'):
raise ValueError(
'Attr \'pad_mode\' of \'Conv1d\' Op passed ' + str(pad_mode) +
', should be one of values in \'valid\', \'same\', \'pad\'.')
self.expand_dims = P.ExpandDims()
self.squeeze = P.Squeeze(2)
self.shape = P.Shape()
def __init__(self,
save_checkpoint_steps=1,
save_checkpoint_seconds=0,
keep_checkpoint_max=5,
keep_checkpoint_per_n_minutes=0,
integrated_save=True,
async_save=False,
saved_network=None,
enc_key=None,
enc_mode='AES-GCM'):
if save_checkpoint_steps is not None:
save_checkpoint_steps = Validator.check_non_negative_int(
save_checkpoint_steps)
if save_checkpoint_seconds is not None:
save_checkpoint_seconds = Validator.check_non_negative_int(
save_checkpoint_seconds)
if keep_checkpoint_max is not None:
keep_checkpoint_max = Validator.check_non_negative_int(
keep_checkpoint_max)
if keep_checkpoint_per_n_minutes is not None:
keep_checkpoint_per_n_minutes = Validator.check_non_negative_int(
keep_checkpoint_per_n_minutes)
if saved_network is not None and not isinstance(
saved_network, nn.Cell):
raise TypeError(
f"The type of saved_network must be None or Cell, but got {str(type(saved_network))}."
)
if not save_checkpoint_steps and not save_checkpoint_seconds and \
not keep_checkpoint_max and not keep_checkpoint_per_n_minutes:
raise ValueError("The input_param can't be all None or 0")
self._save_checkpoint_steps = save_checkpoint_steps
self._save_checkpoint_seconds = save_checkpoint_seconds
if self._save_checkpoint_steps and self._save_checkpoint_steps > 0:
self._save_checkpoint_seconds = None
self._keep_checkpoint_max = keep_checkpoint_max
self._keep_checkpoint_per_n_minutes = keep_checkpoint_per_n_minutes
if self._keep_checkpoint_max and self._keep_checkpoint_max > 0:
self._keep_checkpoint_per_n_minutes = None
else:
if not self._keep_checkpoint_per_n_minutes or self._keep_checkpoint_per_n_minutes == 0:
self._keep_checkpoint_max = 1
self._integrated_save = Validator.check_bool(integrated_save)
self._async_save = Validator.check_bool(async_save)
self._saved_network = saved_network
self._enc_key = Validator.check_isinstance('enc_key', enc_key,
(type(None), bytes))
self._enc_mode = Validator.check_isinstance('enc_mode', enc_mode, str)
def _get_global_and_op_seed():
"""Get global_seed and op_seed."""
global_seed = get_seed()
op_seed = get_seed()
if global_seed == 0:
global_seed = DEFAULT_GRAPH_SEED
elif global_seed is None:
global_seed = 0
if op_seed is None:
op_seed = 0
Validator.check_non_negative_int(op_seed, "seed", "init")
temp_seed = _get_op_seed(op_seed, "init")
seeds = _truncate_seed(global_seed), _truncate_seed(temp_seed)
return seeds
def _get_graph_seed(op_seed, kernel_name):
"""
Get the graph-level seed.
Graph-level seed is used as a global variable, that can be used in different ops in case op-level seed is not set.
If op-level seed is 0, use graph-level seed; if graph-level seed is also 0, the system would generate a
random seed.
Note:
For each seed, either op-seed or graph-seed, a random sequence will be generated relating to this seed.
So, the state of the seed regarding to this op should be recorded.
A simple illustration should be:
If a random op is called twice within one program, the two results should be different:
minval = Tensor(1.0, mstype.float32)
maxval = Tensor(2.0, mstype.float32)
print(C.uniform((1, 4), minval, maxval, seed=1)) # generates 'A1'
print(C.uniform((1, 4), minval, maxval, seed=1)) # generates 'A2'
If the same program runs again, it repeat the results:
print(C.uniform((1, 4), minval, maxval, seed=1)) # generates 'A1'
print(C.uniform((1, 4), minval, maxval, seed=1)) # generates 'A2'
Returns:
Integer. The current graph-level seed.
Examples:
>>> print(_get_graph_seed(0, 'normal'))
(0, 0)
"""
global_seed = get_seed()
if global_seed == 0:
global_seed = DEFAULT_GRAPH_SEED
elif global_seed is None:
global_seed = 0
if op_seed is None:
op_seed = 0
# neither global seed or op seed is set, return (0, 0) to let kernel choose random seed.
if global_seed == 0 and op_seed == 0:
seeds = 0, 0
else:
Validator.check_non_negative_int(op_seed, "seed", kernel_name)
temp_seed = _get_op_seed(op_seed, kernel_name)
seeds = _truncate_seed(global_seed), _truncate_seed(temp_seed)
_update_seeds(op_seed, kernel_name)
return seeds
def __init__(self,
save_checkpoint_steps=1,
save_checkpoint_seconds=0,
keep_checkpoint_max=5,
keep_checkpoint_per_n_minutes=0,
integrated_save=True,
async_save=False):
if save_checkpoint_steps is not None:
save_checkpoint_steps = Validator.check_non_negative_int(
save_checkpoint_steps)
if save_checkpoint_seconds is not None:
save_checkpoint_seconds = Validator.check_non_negative_int(
save_checkpoint_seconds)
if keep_checkpoint_max is not None:
keep_checkpoint_max = Validator.check_non_negative_int(
keep_checkpoint_max)
if keep_checkpoint_per_n_minutes is not None:
keep_checkpoint_per_n_minutes = Validator.check_non_negative_int(
keep_checkpoint_per_n_minutes)
if not save_checkpoint_steps and not save_checkpoint_seconds and \
not keep_checkpoint_max and not keep_checkpoint_per_n_minutes:
raise ValueError("The input_param can't be all None or 0")
self._save_checkpoint_steps = save_checkpoint_steps
self._save_checkpoint_seconds = save_checkpoint_seconds
if self._save_checkpoint_steps and self._save_checkpoint_steps > 0:
self._save_checkpoint_seconds = None
self._keep_checkpoint_max = keep_checkpoint_max
self._keep_checkpoint_per_n_minutes = keep_checkpoint_per_n_minutes
if self._keep_checkpoint_max and self._keep_checkpoint_max > 0:
self._keep_checkpoint_per_n_minutes = None
else:
if not self._keep_checkpoint_per_n_minutes or self._keep_checkpoint_per_n_minutes == 0:
self._keep_checkpoint_max = 1
self._integrated_save = Validator.check_bool(integrated_save)
self._async_save = Validator.check_bool(async_save)
def _check_axes_for_batch_dot(x1_shape, x2_shape, axes):
"""
Check whether axes are valid and cast axes from tuple to list
"""
if axes is None:
if len(x2_shape) == 2:
axes = [len(x1_shape) - 1, len(x2_shape) - 1]
else:
axes = [len(x1_shape) - 1, len(x2_shape) - 2]
if isinstance(axes, (list, tuple)):
if 0 in axes:
raise ValueError("Batch dim cannot be used as in axes.")
if len(axes) != 2:
raise ValueError("Require two axes inputs, given less")
if isinstance(axes, tuple):
axes = list(axes)
validator.check_value_type('axes[0]', axes[0], [int], 'batch_dot')
validator.check_value_type('axes[1]', axes[1], [int], 'batch_dot')
# Reverse if axis < 0
if axes[0] < 0:
axes[0] += len(x1_shape)
if axes[1] < 0:
axes[1] += len(x2_shape)
validator.check_non_negative_int(axes[0], 'reversed axes[0]',
'batch_dot')
validator.check_non_negative_int(axes[1], 'reversed axes[1]',
'batch_dot')
if axes[0] > len(x1_shape) or axes[1] > len(x2_shape):
raise ValueError(
"Axes value too high for given input arrays dimensions.")
elif isinstance(axes, int):
if axes == 0:
raise ValueError("Batch dim cannot be used as in axes.")
if axes < 0:
axes = [axes + len(x1_shape), axes + len(x2_shape)]
validator.check_non_negative_int(axes[0], 'reversed axes',
'batch_dot')
elif axes > len(x1_shape) or axes > len(x2_shape):
raise ValueError(
"Axes value too high for given input arrays dimensions.")
else:
axes = [axes, axes]
else:
raise ValueError(
"Axes type must be one of those: int, tuple(int), list(int).")
return axes
def set_seed(seed):
"""
Set global random seed.
Note:
The global seed is used by numpy.random, mindspore.common.Initializer, mindspore.ops.composite.random_ops and
mindspore.nn.probability.distribution.
If global seed is not set, these packages will use their own default seed independently, numpy.random and
mindspore.common.Initializer will choose a random seed, mindspore.ops.composite.random_ops and
mindspore.nn.probability.distribution will use zero.
Seed set by numpy.random.seed() only used by numpy.random, while seed set by this API will also used by
numpy.random, so just set all seed by this API is recommended.
Args:
seed (int): The seed to be set.
Raises:
ValueError: If seed is invalid (< 0).
TypeError: If seed isn't a int.
Examples:
>>> from mindspore.ops import composite as C
>>> from mindspore import Tensor
>>>
>>> # Note: (1) Please make sure the code is running in PYNATIVE MODE;
>>> # (2) Because Composite-level ops need parameters to be Tensors, for below examples,
>>> # when using C.uniform operator, minval and maxval are initialised as:
>>> minval = Tensor(1.0, ms.float32)
>>> maxval = Tensor(2.0, ms.float32)
>>>
>>> # 1. If global seed is not set, numpy.random and initializer will choose a random seed:
>>> np_1 = np.random.normal(0, 1, [1]).astype(np.float32) # A1
>>> np_1 = np.random.normal(0, 1, [1]).astype(np.float32) # A2
>>> w1 = Parameter(initializer("uniform", [2, 2], ms.float32), name="w1") # W1
>>> w1 = Parameter(initializer("uniform", [2, 2], ms.float32), name="w1") # W2
>>> # Rerun the program will get different results:
>>> np_1 = np.random.normal(0, 1, [1]).astype(np.float32) # A3
>>> np_1 = np.random.normal(0, 1, [1]).astype(np.float32) # A4
>>> w1 = Parameter(initializer("uniform", [2, 2], ms.float32), name="w1") # W3
>>> w1 = Parameter(initializer("uniform", [2, 2], ms.float32), name="w1") # W4
>>>
>>> # 2. If global seed is set, numpy.random and initializer will use it:
>>> set_seed(1234)
>>> np_1 = np.random.normal(0, 1, [1]).astype(np.float32) # A1
>>> np_1 = np.random.normal(0, 1, [1]).astype(np.float32) # A2
>>> w1 = Parameter(initializer("uniform", [2, 2], ms.float32), name="w1") # W1
>>> w1 = Parameter(initializer("uniform", [2, 2], ms.float32), name="w1") # W2
>>> # Rerun the program will get the same results:
>>> set_seed(1234)
>>> np_1 = np.random.normal(0, 1, [1]).astype(np.float32) # A1
>>> np_1 = np.random.normal(0, 1, [1]).astype(np.float32) # A2
>>> w1 = Parameter(initializer("uniform", [2, 2], ms.float32), name="w1") # W1
>>> w1 = Parameter(initializer("uniform", [2, 2], ms.float32), name="w1") # W2
>>>
>>> # 3. If neither global seed nor op seed is set, mindspore.ops.composite.random_ops and
>>> # mindspore.nn.probability.distribution will choose a random seed:
>>> c1 = C.uniform((1, 4), minval, maxval) # C1
>>> c2 = C.uniform((1, 4), minval, maxval) # C2
>>> # Rerun the program will get different results:
>>> c1 = C.uniform((1, 4), minval, maxval) # C3
>>> c2 = C.uniform((1, 4), minval, maxval) # C4
>>>
>>> # 4. If global seed is set, but op seed is not set, mindspore.ops.composite.random_ops and
>>> # mindspore.nn.probability.distribution will calculate a seed according to global seed and
>>> # default op seed. Each call will change the default op seed, thus each call get different
>>> # results.
>>> set_seed(1234)
>>> c1 = C.uniform((1, 4), minval, maxval) # C1
>>> c2 = C.uniform((1, 4), minval, maxval) # C2
>>> # Rerun the program will get the same results:
>>> set_seed(1234)
>>> c1 = C.uniform((1, 4), minval, maxval) # C1
>>> c2 = C.uniform((1, 4), minval, maxval) # C2
>>>
>>> # 5. If both global seed and op seed are set, mindspore.ops.composite.random_ops and
>>> # mindspore.nn.probability.distribution will calculate a seed according to global seed and
>>> # op seed counter. Each call will change the op seed counter, thus each call get different
>>> # results.
>>> set_seed(1234)
>>> c1 = C.uniform((1, 4), minval, maxval, seed=2) # C1
>>> c2 = C.uniform((1, 4), minval, maxval, seed=2) # C2
>>> # Rerun the program will get the same results:
>>> set_seed(1234)
>>> c1 = C.uniform((1, 4), minval, maxval, seed=2) # C1
>>> c2 = C.uniform((1, 4), minval, maxval, seed=2) # C2
>>>
>>> # 6. If op seed is set but global seed is not set, 0 will be used as global seed. Then
>>> # mindspore.ops.composite.random_ops and mindspore.nn.probability.distribution act as in
>>> # condition 5.
>>> c1 = C.uniform((1, 4), minval, maxval, seed=2) # C1
>>> c2 = C.uniform((1, 4), minval, maxval, seed=2) # C2
>>> # Rerun the program will get the same results:
>>> c1 = C.uniform((1, 4), minval, maxval, seed=2) # C1
>>> c2 = C.uniform((1, 4), minval, maxval, seed=2) # C2
>>>
>>> # 7. Recall set_seed() in the program will reset numpy seed and op seed counter of
>>> # mindspore.ops.composite.random_ops and mindspore.nn.probability.distribution.
>>> set_seed(1234)
>>> np_1 = np.random.normal(0, 1, [1]).astype(np.float32) # A1
>>> c1 = C.uniform((1, 4), minval, maxval, seed=2) # C1
>>> set_seed(1234)
>>> np_2 = np.random.normal(0, 1, [1]).astype(np.float32) # still get A1
>>> c2 = C.uniform((1, 4), minval, maxval, seed=2) # still get C1
"""
if not isinstance(seed, int):
raise TypeError("The seed must be type of int.")
Validator.check_non_negative_int(seed, "seed", "global_seed")
# We put import here to solve an ut/cpp core issue
import mindspore.dataset as de
np.random.seed(seed)
de.config.set_seed(seed)
_reset_op_seed()
global _GLOBAL_SEED
_GLOBAL_SEED = seed
def __init__(self,
vocab_size,
embedding_size,
param_init='normal',
target='CPU',
slice_mode='batch_slice',
manual_shapes=None,
max_norm=None,
sparse=True,
vocab_cache_size=0):
super(EmbeddingLookup, self).__init__()
validator.check_value_type('sparse', sparse, [bool], self.cls_name)
self.vocab_size = validator.check_positive_int(vocab_size,
'vocab_size')
self.vocab_cache_size = validator.check_non_negative_int(
vocab_cache_size, 'vocab_cache_size')
self.target = target
self.sparse = sparse
self.cache_enable = self.vocab_cache_size > 0
self.forward_unique = False
if target not in ('CPU', 'DEVICE'):
raise ValueError(
'Attr \'target\' of \'EmbeddingLookup\' Op passed ' +
str(target) +
', should be one of values in \'CPU\', \'DEVICE\'.')
if not sparse and target == 'CPU':
raise ValueError(
'When target is CPU, embedding_lookup must be sparse.')
if sparse:
self.gatherv2 = P.SparseGatherV2()
else:
self.gatherv2 = P.Gather()
self.embeddinglookup = P.EmbeddingLookup().add_prim_attr(
'primitive_target', 'CPU')
enable_ps = _get_ps_context("enable_ps")
if enable_ps:
self._process_vocab_cache(slice_mode)
self.embedding_size = validator.check_positive_int(
embedding_size, 'embedding_size')
self.embedding_table = Parameter(initializer(
param_init, [self.vocab_size, self.embedding_size]),
name='embedding_table')
parallel_mode = _get_parallel_mode()
is_auto_parallel = parallel_mode in (ParallelMode.SEMI_AUTO_PARALLEL,
ParallelMode.AUTO_PARALLEL)
self.gather_revert = P.Gather()
self.reshape_first = P.Reshape()
self.reshape = P.Reshape()
self.unique = P.Unique()
self.shape = P.Shape()
if is_auto_parallel:
self.unique = P.Unique().shard(((1, ), ))
if self.cache_enable and enable_ps:
self._set_voacb_cache_enable_for_ps(vocab_cache_size,
embedding_size, vocab_size)
if is_auto_parallel:
self.unique.add_prim_attr('cache_enable', True)
indices_shape_size = 2
if slice_mode == "field_slice" and is_auto_parallel:
if not manual_shapes:
raise ValueError(
"in slice field mode, the manual_shapes should not be none"
)
if not isinstance(manual_shapes, tuple):
raise TypeError(
"manual_shapes type must be tuple(int) cannot be {}!".
format(type(manual_shapes)))
for dim in manual_shapes:
validator.check_positive_int(dim, 'manual shape dim',
self.cls_name)
self.gatherv2.add_prim_attr("manual_split", manual_shapes)
self.embeddinglookup.add_prim_attr("manual_split", manual_shapes)
self.gatherv2.shard(((get_group_size(), 1), (1, get_group_size())))
self.embeddinglookup.shard(
((get_group_size(), 1), (1, get_group_size())))
elif slice_mode == "table_row_slice" and is_auto_parallel:
full_batch = _get_full_batch()
if (target == 'DEVICE'
and not full_batch) or (self.cache_enable and enable_ps
and sparse):
indices_shape_size = 1
self.gather_revert.shard(((1, 1), (get_group_size(), )))
self.forward_unique = True
indices_strategy = (1, ) * indices_shape_size
self.gatherv2.shard(((get_group_size(), 1), indices_strategy))
self.embeddinglookup.shard(
((get_group_size(), 1), indices_strategy))
elif slice_mode == "table_column_slice" and is_auto_parallel:
if target == 'DEVICE':
indices_shape_size = 1
self.gather_revert.shard(((1, get_group_size()), (1, )))
self.forward_unique = True
indices_strategy = (1, ) * indices_shape_size
self.gatherv2.shard(((1, get_group_size()), indices_strategy))
self.embeddinglookup.shard(
((1, get_group_size()), indices_strategy))
elif slice_mode == "batch_slice" and is_auto_parallel:
indices_strategy = [get_group_size()]
indices_strategy.extend([1] * (indices_shape_size - 1))
indices_strategy = tuple(indices_strategy)
self.gatherv2.shard(((1, 1), indices_strategy))
self.embeddinglookup.shard(((1, 1), indices_strategy))
else:
if is_auto_parallel:
raise ValueError(
"slice_mode should support mode in nn.EmbeddingLookup, but get "
+ str(slice_mode))
if self.cache_enable and not enable_ps:
if parallel_mode != ParallelMode.STAND_ALONE:
raise ValueError(
"parallel mode haven't supported cache enable yet.")
self._set_cache_enable()
self.embedding_table.unique = self.forward_unique
self.max_norm = max_norm
if self.max_norm is not None:
self.max_norm = validator.check_positive_float(
self.max_norm, 'max_norm', self.cls_name)
self.max_norm = Tensor(self.max_norm, dtype=mstype.float32)
def __init__(self,
in_channels,
out_channels,
kernel_size,
stride=1,
pad_mode='same',
padding=0,
dilation=1,
group=1,
has_bias=False,
weight_init='normal',
bias_init='zeros'):
Validator.check_value_type("kernel_size", kernel_size, [int],
self.cls_name)
Validator.check_value_type("stride", stride, [int], self.cls_name)
Validator.check_value_type("padding", padding, [int], self.cls_name)
Validator.check_value_type("dilation", dilation, [int], self.cls_name)
Validator.check_int(kernel_size, 1, Rel.GE, 'kernel_size',
self.cls_name)
Validator.check_int(stride, 1, Rel.GE, 'stride', self.cls_name)
Validator.check_non_negative_int(padding, 'padding', self.cls_name)
Validator.check_int(dilation, 1, Rel.GE, 'dilation', self.cls_name)
kernel_size = (1, kernel_size)
stride = (1, stride)
dilation = (1, dilation)
get_shape = P.Shape()
get_dtype = P.DType()
if isinstance(weight_init, Tensor):
weight_init_shape = get_shape(weight_init)
Validator.check_equal_int(len(weight_init_shape), 3,
'weight_init_shape', self.cls_name)
weight_init_dtype = get_dtype(weight_init)
weight_init_value = weight_init.asnumpy()
weight_init_value = np.expand_dims(weight_init_value, 2)
weight_init = Tensor(weight_init_value, weight_init_dtype)
# out_channels and in_channels swap.
# cause Conv2DBackpropInput's out_channel refers to Conv2D's out_channel,
# then Conv1dTranspose's out_channel refers to Conv2DBackpropInput's in_channel.
super(Conv1dTranspose, self).__init__(in_channels,
out_channels,
kernel_size,
stride,
pad_mode,
padding,
dilation,
group,
has_bias,
weight_init,
bias_init,
transposed=True)
self.padding = (0, 0, padding, padding)
self.in_channels = in_channels
self.out_channels = out_channels
self.shape = P.Shape()
if pad_mode not in ('valid', 'same', 'pad'):
raise ValueError(
'Attr \'pad_mode\' of \'Conv1dTranspose\' Op passed ' +
str(pad_mode) +
', should be one of values in \'valid\', \'same\', \'pad\'.')
self.is_valid = self.pad_mode == 'valid'
self.is_same = self.pad_mode == 'same'
self.is_pad = self.pad_mode == 'pad'
if Validator.check_bool(has_bias):
self.bias = Parameter(initializer(bias_init, [out_channels]),
name='bias')
# cause Conv2DBackpropInput's out_channel refers to Conv2D's out_channel.
self.conv2d_transpose = P.Conv2DBackpropInput(out_channel=in_channels,
kernel_size=kernel_size,
mode=1,
pad_mode=pad_mode,
pad=self.padding,
stride=stride,
dilation=dilation,
group=group)
self.bias_add = P.BiasAdd()
self.expand_dims = P.ExpandDims()
self.squeeze = P.Squeeze(2)
def __init__(self,
in_channels,
out_channels,
kernel_size,
stride,
pad_mode,
padding,
dilation,
group,
has_bias,
weight_init,
bias_init,
data_format='NCHW',
transposed=False):
super(_Conv, self).__init__()
self.in_channels = Validator.check_positive_int(in_channels)
self.out_channels = Validator.check_positive_int(out_channels)
self.kernel_size = kernel_size
self.stride = stride
self.pad_mode = pad_mode
self.weight_init = weight_init
self.bias_init = bias_init
self.format = Validator.check_string(data_format, ['NCHW', 'NHWC'],
'format', self.cls_name)
if context.get_context(
"device_target") != "GPU" and self.format == "NHWC":
raise ValueError("NHWC format only support in GPU target.")
if isinstance(padding, int):
Validator.check_non_negative_int(padding, 'padding', self.cls_name)
self.padding = padding
elif isinstance(padding, tuple):
for pad in padding:
Validator.check_non_negative_int(pad, 'padding item',
self.cls_name)
self.padding = padding
else:
raise TypeError(
"padding type must be int/tuple(int) cannot be {}!".format(
type(padding)))
self.dilation = dilation
self.group = Validator.check_positive_int(group)
self.has_bias = has_bias
if (not isinstance(kernel_size[0], int)) or (not isinstance(kernel_size[1], int)) or \
isinstance(kernel_size[0], bool) or isinstance(kernel_size[1], bool) or \
kernel_size[0] < 1 or kernel_size[1] < 1:
raise ValueError(
"Attr 'kernel_size' of 'Conv2D' Op passed " +
str(self.kernel_size) +
", should be a int or tuple and equal to or greater than 1.")
if (not isinstance(stride[0], int)) or (not isinstance(stride[1], int)) or \
isinstance(stride[0], bool) or isinstance(stride[1], bool) or stride[0] < 1 or stride[1] < 1:
raise ValueError(
"Attr 'stride' of 'Conv2D' Op passed " + str(self.stride) +
", should be a int or tuple and equal to or greater than 1.")
if (not isinstance(dilation[0], int)) or (not isinstance(dilation[1], int)) or \
isinstance(dilation[0], bool) or isinstance(dilation[1], bool) or dilation[0] < 1 or dilation[1] < 1:
raise ValueError(
"Attr 'dilation' of 'Conv2D' Op passed " + str(self.dilation) +
", should be a int or tuple and equal to or greater than 1.")
if in_channels % group != 0:
raise ValueError(
"Attr 'in_channels' of 'Conv2D' Op must be divisible by "
"attr 'group' of 'Conv2D' Op.")
if out_channels % group != 0:
raise ValueError(
"Attr 'out_channels' of 'Conv2D' Op must be divisible by "
"attr 'group' of 'Conv2D' Op.")
if transposed:
shape = [in_channels, out_channels // group, *kernel_size]
else:
shape = [out_channels, in_channels // group, *kernel_size] if self.format == "NCHW" else \
[out_channels, *kernel_size, in_channels // group]
self.weight = Parameter(initializer(self.weight_init, shape),
name='weight')
if Validator.check_bool(has_bias):
self.bias = Parameter(initializer(self.bias_init, [out_channels]),
name='bias')
else:
if self.bias_init != 'zeros':
logger.warning(
"Value of 'has_bias' is False, value of 'bias_init' will be ignored."
)
self.bias = None
def set_seed(seed):
"""
Set global random seed.
Note:
The global seed is used by numpy.random, mindspore.common.Initializer, mindspore.ops.composite.random_ops and
mindspore.nn.probability.distribution.
If global seed is not set, these packages will use their own default seed independently, numpy.random and
mindspore.common.Initializer will choose a random seed, mindspore.ops.composite.random_ops and
mindspore.nn.probability.distribution will use zero.
Seed set by numpy.random.seed() only used by numpy.random, while seed set by this API will also used by
numpy.random, so just set all seed by this API is recommended.
Args:
seed (int): The seed to be set.
Raises:
ValueError: If seed is invalid (< 0).
TypeError: If seed isn't a int.
Examples:
>>> # 1. If global seed is not set, numpy.random and initializer will choose a random seed:
>>> np_1 = np.random.normal(0, 1, [1]).astype(np.float32) # A1
>>> np_1 = np.random.normal(0, 1, [1]).astype(np.float32) # A2
>>> w1 = Parameter(initializer("uniform", [2, 2], ms.float32), name="w1") # W1
>>> w1 = Parameter(initializer("uniform", [2, 2], ms.float32), name="w1") # W2
>>> # Rerun the program will get diferent results:
>>> np_1 = np.random.normal(0, 1, [1]).astype(np.float32) # A3
>>> np_1 = np.random.normal(0, 1, [1]).astype(np.float32) # A4
>>> w1 = Parameter(initializer("uniform", [2, 2], ms.float32), name="w1") # W3
>>> w1 = Parameter(initializer("uniform", [2, 2], ms.float32), name="w1") # W4
>>>
>>> 2. If global seed is set, numpy.random and initializer will use it:
>>> set_seed(1234)
>>> np_1 = np.random.normal(0, 1, [1]).astype(np.float32) # A1
>>> np_1 = np.random.normal(0, 1, [1]).astype(np.float32) # A2
>>> w1 = Parameter(initializer("uniform", [2, 2], ms.float32), name="w1") # W1
>>> w1 = Parameter(initializer("uniform", [2, 2], ms.float32), name="w1") # W2
>>> # Rerun the program will get the same results:
>>> set_seed(1234)
>>> np_1 = np.random.normal(0, 1, [1]).astype(np.float32) # A1
>>> np_1 = np.random.normal(0, 1, [1]).astype(np.float32) # A2
>>> w1 = Parameter(initializer("uniform", [2, 2], ms.float32), name="w1") # W1
>>> w1 = Parameter(initializer("uniform", [2, 2], ms.float32), name="w1") # W2
>>>
>>> # 3. If neither global seed nor op seed is set, mindspore.ops.composite.random_ops and
>>> # mindspore.nn.probability.distribution will choose a random seed:
>>> c1 = C.uniform((1, 4)) # C1
>>> c2 = C.uniform((1, 4)) # C2
>>> Rerun the program will get different results:
>>> c1 = C.uniform((1, 4)) # C3
>>> c2 = C.uniform((1, 4)) # C4
>>>
>>> # 4. If global seed is set, but op seed is not set, mindspore.ops.composite.random_ops and
>>> # mindspore.nn.probability.distribution will caculate a seed according to global seed and
>>> # default op seed. Each call will change the default op seed, thus each call get different
>>> # results.
>>> set_seed(1234)
>>> c1 = C.uniform((1, 4)) # C1
>>> c2 = C.uniform((1, 4)) # C2
>>> # Rerun the program will get the same results:
>>> set_seed(1234)
>>> c1 = C.uniform((1, 4)) # C1
>>> c2 = C.uniform((1, 4)) # C2
>>>
>>> # 5. If both global seed and op seed are set, mindspore.ops.composite.random_ops and
>>> # mindspore.nn.probability.distribution will caculate a seed according to global seed and
>>> # op seed counter. Each call will change the op seed counter, thus each call get different
>>> # results.
>>> set_seed(1234)
>>> c1 = C.uniform((1, 4), seed=2) # C1
>>> c2 = C.uniform((1, 4), seed=2) # C2
>>> Rerun the program will get the same results:
>>> set_seed(1234)
>>> c1 = C.uniform((1, 4), seed=2) # C1
>>> c2 = C.uniform((1, 4), seed=2) # C2
>>>
>>> # 6. If op seed is set but global seed is not set, 0 will be used as global seed. Then
>>> # mindspore.ops.composite.random_ops and mindspore.nn.probability.distribution act as in
>>> # condition 5.
>>> c1 = C.uniform((1, 4), seed=2) # C1
>>> c2 = C.uniform((1, 4), seed=2) # C2
>>> #Rerun the program will get the same results:
>>> c1 = C.uniform((1, 4), seed=2) # C1
>>> c2 = C.uniform((1, 4), seed=2) # C2
>>>
>>> # 7. Recall set_seed() in the program will reset numpy seed and op seed counter of
>>> # mindspore.ops.composite.random_ops and mindspore.nn.probability.distribution.
>>> set_seed(1234)
>>> np_1 = np.random.normal(0, 1, [1]).astype(np.float32) # A1
>>> c1 = C.uniform((1, 4), seed=2) # C1
>>> set_seed(1234)
>>> np_2 = np.random.normal(0, 1, [1]).astype(np.float32) # still get A1
>>> c2 = C.uniform((1, 4), seed=2) # still get C1
"""
if not isinstance(seed, int):
raise TypeError("The seed must be type of int.")
Validator.check_non_negative_int(seed, "seed", "global_seed")
np.random.seed(seed)
de.config.set_seed(seed)
_reset_op_seed()
global _GLOBAL_SEED
_GLOBAL_SEED = seed
