def test_nobroadcast_fp16():
context.set_context(mode=context.GRAPH_MODE, device_target='GPU')
np.random.seed(42)
x1_np = np.random.rand(10, 20).astype(np.float16)
x2_np = np.random.rand(10, 20).astype(np.float16)
output_ms = P.Minimum()(Tensor(x1_np), Tensor(x2_np))
output_np = np.minimum(x1_np, x2_np)
assert np.allclose(output_ms.asnumpy(), output_np)
output_ms = P.Maximum()(Tensor(x1_np), Tensor(x2_np))
output_np = np.maximum(x1_np, x2_np)
assert np.allclose(output_ms.asnumpy(), output_np)
output_ms = P.Greater()(Tensor(x1_np), Tensor(x2_np))
output_np = x1_np > x2_np
assert np.allclose(output_ms.asnumpy(), output_np)
output_ms = P.Less()(Tensor(x1_np), Tensor(x2_np))
output_np = x1_np < x2_np
assert np.allclose(output_ms.asnumpy(), output_np)
output_ms = P.Pow()(Tensor(x1_np), Tensor(x2_np))
output_np = np.power(x1_np, x2_np)
assert np.allclose(output_ms.asnumpy(), output_np)
output_ms = P.RealDiv()(Tensor(x1_np), Tensor(x2_np))
output_np = x1_np / x2_np
assert np.allclose(output_ms.asnumpy(), output_np)
output_ms = P.Mul()(Tensor(x1_np), Tensor(x2_np))
output_np = x1_np * x2_np
assert np.allclose(output_ms.asnumpy(), output_np)
output_ms = P.Sub()(Tensor(x1_np), Tensor(x2_np))
output_np = x1_np - x2_np
assert np.allclose(output_ms.asnumpy(), output_np)
output_ms = P.DivNoNan()(Tensor(x1_np), Tensor(x2_np))
output_np = x1_np / x2_np
assert np.allclose(output_ms.asnumpy(), output_np)
x2_np_zero = np.zeros_like(x2_np)
output_ms = P.DivNoNan()(Tensor(x1_np), Tensor(x2_np_zero))
assert np.allclose(output_ms.asnumpy(), x2_np_zero)
output_ms = P.Mod()(Tensor(x1_np), Tensor(x2_np))
output_np = np.fmod(x1_np, x2_np)
assert np.allclose(output_ms.asnumpy(), output_np)
output_ms = P.FloorMod()(Tensor(x1_np), Tensor(x2_np))
output_np = np.mod(x1_np, x2_np)
assert np.allclose(output_ms.asnumpy(), output_np)
def test_broadcast_diff_dims():
context.set_context(mode=context.GRAPH_MODE, device_target='GPU')
np.random.seed(42)
x1_np = np.random.rand(2).astype(np.float32)
x2_np = np.random.rand(2, 1).astype(np.float32)
x1_np_int32 = np.random.randint(0, 100, (2)).astype(np.int32)
x2_np_int32 = np.random.randint(0, 100, (2, 1)).astype(np.int32)
output_ms = P.Minimum()(Tensor(x1_np), Tensor(x2_np))
output_np = np.minimum(x1_np, x2_np)
assert np.allclose(output_ms.asnumpy(), output_np)
output_ms = P.Maximum()(Tensor(x1_np), Tensor(x2_np))
output_np = np.maximum(x1_np, x2_np)
assert np.allclose(output_ms.asnumpy(), output_np)
output_ms = P.Greater()(Tensor(x1_np_int32), Tensor(x2_np_int32))
output_np = x1_np_int32 > x2_np_int32
assert np.allclose(output_ms.asnumpy(), output_np)
output_ms = P.Greater()(Tensor(x1_np), Tensor(x2_np))
output_np = x1_np > x2_np
assert np.allclose(output_ms.asnumpy(), output_np)
output_ms = P.Less()(Tensor(x1_np), Tensor(x2_np))
output_np = x1_np < x2_np
assert np.allclose(output_ms.asnumpy(), output_np)
output_ms = P.Less()(Tensor(x1_np_int32), Tensor(x2_np_int32))
output_np = x1_np_int32 < x2_np_int32
assert np.allclose(output_ms.asnumpy(), output_np)
output_ms = P.Pow()(Tensor(x1_np), Tensor(x2_np))
output_np = np.power(x1_np, x2_np)
assert np.allclose(output_ms.asnumpy(), output_np)
output_ms = P.RealDiv()(Tensor(x1_np), Tensor(x2_np))
output_np = x1_np / x2_np
assert np.allclose(output_ms.asnumpy(), output_np)
output_ms = P.Mul()(Tensor(x1_np), Tensor(x2_np))
output_np = x1_np * x2_np
assert np.allclose(output_ms.asnumpy(), output_np)
output_ms = P.Sub()(Tensor(x1_np), Tensor(x2_np))
output_np = x1_np - x2_np
assert np.allclose(output_ms.asnumpy(), output_np)
output_ms = P.DivNoNan()(Tensor(x1_np), Tensor(x2_np))
output_np = x1_np / x2_np
assert np.allclose(output_ms.asnumpy(), output_np)
x2_np_zero = np.zeros_like(x2_np)
output_ms = P.DivNoNan()(Tensor(x1_np), Tensor(x2_np_zero))
assert np.allclose(output_ms.asnumpy(), x2_np_zero)
def test_divnonan_uint8():
context.set_context(mode=context.GRAPH_MODE, device_target='GPU')
np.random.seed(42)
x1_np_uint8 = np.random.randint(1, 100, (10, 20)).astype(np.uint8)
x2_np_uint8 = np.random.randint(1, 100, (10, 20)).astype(np.uint8)
output_ms = P.DivNoNan()(Tensor(x1_np_uint8), Tensor(x2_np_uint8))
output_np = x1_np_uint8 // x2_np_uint8
print(output_ms.asnumpy(), output_np)
assert np.allclose(output_ms.asnumpy(), output_np)
x2_np_zero = np.zeros_like(x2_np_uint8)
output_ms = P.DivNoNan()(Tensor(x1_np_uint8), Tensor(x2_np_zero))
assert np.allclose(output_ms.asnumpy(), x2_np_zero)
def __init__(self,
vocab_size,
embedding_size,
field_size,
param_init='normal',
target='CPU',
slice_mode='batch_slice',
feature_num_list=None,
max_norm=None,
sparse=True,
operator='SUM'):
super(MultiFieldEmbeddingLookup,
self).__init__(vocab_size, embedding_size, param_init, target,
slice_mode, feature_num_list, max_norm, sparse)
self.field_size = validator.check_positive_int(field_size,
'field_size')
self.operator = operator
self.mul = P.Mul()
self.inf_mask_mul = P.Mul()
self.bias_add = P.Add()
self.inf_add = P.Add()
self.merge_op = None
self.count_op = P.UnsortedSegmentSum()
self.abs = P.Abs()
self.equal = P.Equal()
self.add = P.Add()
self.cast = P.Cast()
self.div_no_nan = P.DivNoNan()
self.expand = P.ExpandDims()
self.max_mask_mul = P.Mul()
self.max_no_equal = P.NotEqual()
if operator == MultiFieldEmbeddingLookup.OPERATOR_SUM:
self.merge_op = P.UnsortedSegmentSum()
elif operator == MultiFieldEmbeddingLookup.OPERATOR_MAX:
self.merge_op = P.UnsortedSegmentMax()
elif operator == MultiFieldEmbeddingLookup.OPERATOR_MEAN:
self.merge_op = P.UnsortedSegmentSum()
else:
raise ValueError(
"The operator supports ['SUM', 'MAX', 'MEAN'], but found: " +
str(operator))
parallel_mode = _get_parallel_mode()
is_auto_parallel = parallel_mode in (ParallelMode.SEMI_AUTO_PARALLEL,
ParallelMode.AUTO_PARALLEL)
if slice_mode in ["table_row_slice", "batch_slice"
] and is_auto_parallel:
self.merge_op.shard(
((get_group_size(), 1, 1), (get_group_size(), 1)))
self.expand.shard(((get_group_size(), ), ))
self.bias_add.shard(((1, 1), (1, 1)))
self.mul.shard(
((get_group_size(), 1, 1), (get_group_size(), 1, 1)))
self.count_op.shard(((get_group_size(), 1), (get_group_size(), 1)))
self.add.shard(((get_group_size(), ), (get_group_size(), )))
self.div_no_nan.shard(
((get_group_size(), 1), (get_group_size(), 1)))
self.max_mask_mul.shard(
((get_group_size(), 1), (get_group_size(), 1)))
self.max_no_equal.shard(((1, ), ()))
if operator == MultiFieldEmbeddingLookup.OPERATOR_MAX:
self.equal.shard(((get_group_size(), 1, 1), ()))
self.inf_mask_mul.shard(((get_group_size(), 1, 1), ()))
self.merge_op.shard(
((get_group_size(), 1), (get_group_size(), )))
self.count_op.shard(
((get_group_size(), ), (get_group_size(), )))
self.inf_add.shard(
((get_group_size(), 1, 1), (get_group_size(), 1, 1)))
elif slice_mode == "table_column_slice" and is_auto_parallel:
self.merge_op.shard(((1, 1, get_group_size()), (1, 1)))
self.div_no_nan.shard(((1, get_group_size()), (1, 1)))
self.bias_add.shard(((1, 1), (1, 1)))
self.mul.shard(((1, 1, 1), (1, 1, get_group_size())))
self.count_op.shard(((1, 1), (1, 1)))
self.add.shard(((1, ), (1, )))
self.max_mask_mul.shard(((1, get_group_size()), (1, 1)))
self.expand.shard(((1, ), ))
self.max_no_equal.shard(((1, ), ()))
if operator == MultiFieldEmbeddingLookup.OPERATOR_MAX:
self.equal.shard(((1, 1, 1), ()))
self.inf_mask_mul.shard(((1, 1, 1), ()))
self.merge_op.shard(((1, get_group_size()), (1, )))
self.count_op.shard(((1, ), (1, )))
self.inf_add.shard(((1, 1, get_group_size()), (1, 1, 1)))
else:
if is_auto_parallel:
raise ValueError(
"slice_mode should be ['table_row_slice', 'batch_slice' and \
'table_column_slice'], but get " + str(slice_mode))
# Min value for fp32
self.negative_inf_value = -3.402823466E+38
def __init__(self, strategy1, strategy2):
super().__init__()
self.matmul = P.MatMul().shard(strategy1)
self.divNoNan = P.DivNoNan().shard(strategy2)
