def __init__(self):
super().__init__()
self.virtual_dataset = _VirtualDataset()
self.matmul1 = P.MatMul()
self.matmul2 = P.MatMul()
self.gelu = P.Gelu()
self.bn1 = bn_with_initialize(2048)
def __init__(self, strategy1, strategy2, strategy3, strategy4, strategy5, strategy6):
super().__init__()
self.matmul1 = P.MatMul().set_strategy(strategy1)
self.matmul2 = P.MatMul().set_strategy(strategy2)
self.gelu = P.Gelu().set_strategy(strategy3)
self.tanh = P.Tanh().set_strategy(strategy4)
self.softmax = P.Softmax(axis=(0, 1)).set_strategy(strategy5)
self.logsoftmax = P.LogSoftmax().set_strategy(strategy6)
def __init__(self, strategy1, strategy2, strategy3):
super().__init__()
self.matmul1 = P.MatMul().shard(strategy1)
self.matmul2 = P.MatMul().shard(strategy2)
self.gelu = P.Gelu().shard(strategy3)
self.tanh = P.Tanh().shard(strategy3)
self.softmax = P.Softmax().shard(strategy3)
self.logsoftmax = P.LogSoftmax().shard(strategy3)
def __init__(self,
in_channels=786,
out_channels=768,
hidden_size=3072,
hidden_dropout=0.1):
super(FeedForward, self).__init__()
self.c_fc = Conv1D(in_channels, hidden_size)
self.c_proj = Conv1D(hidden_size, out_channels)
self.layernorm = LayerNorm(in_channels=in_channels)
self.residual_connect = ResidualConnection(dropout_prob=hidden_dropout)
self.gelu_act = P.Gelu()
self.dropout = nn.Dropout(1 - hidden_dropout)
self.use_dropout = hidden_dropout > 0
self.reshape = P.Reshape()
def __init__(self, strategy1, strategy2, strategy3):
super().__init__()
self.matmul1 = P.MatMul().set_strategy(strategy1)
self.matmul2 = P.MatMul().set_strategy(strategy2)
self.gelu = P.Gelu().set_strategy(strategy3)
def __init__(self, strategy0, strategy1, strategy2, strategy3):
super().__init__()
self.virtual_dataset = _VirtualDataset().set_strategy(strategy0)
self.matmul1 = P.MatMul().set_strategy(strategy1)
self.matmul2 = P.MatMul().set_strategy(strategy2)
self.gelu = P.Gelu().set_strategy(strategy3)
def __init__(self):
super(GeluNet, self).__init__()
self.gelu = P.Gelu()
def __init__(self):
super(GELU, self).__init__()
self.matmul = P.MatMul()
self.gelu = P.Gelu()
def __init__(self):
super(MEGeluLargeIn, self).__init__()
self.matmul = P.MatMul()
self.gelu = P.Gelu()
def __init__(self, strategy1, strategy2):
super().__init__()
self.matmul = P.MatMul().shard(strategy1)
self.gelu = P.Gelu().shard(strategy2)
'desc_inputs': [Tensor(np.array([0, 1]).astype(np.float32)),
Tensor(np.array([1, 1]).astype(np.float32))],
'desc_bprop': [[2]]})
]
test_case_nn_ops = [
('BiasAdd', {
'block': P.BiasAdd(),
'desc_inputs': [[1, 3, 3, 3], [3]],
'desc_bprop': [[1, 3, 3, 3]]}),
('BiasAddGrad', {
'block': G.BiasAddGrad(),
'desc_inputs': [[1, 3, 3, 3]],
'skip': ['backward']}),
('Gelu', {
'block': P.Gelu(),
'desc_inputs': [[1, 3, 4, 4]],
'desc_bprop': [[1, 3, 4, 4]]}),
('GeluGrad', {
'block': G.GeluGrad(),
'desc_inputs': [[2, 2], [2, 2], [2, 2]],
'desc_bprop': [[2, 2]],
'skip': ['backward']}),
('Tanh', {
'block': P.Tanh(),
'desc_inputs': [[1, 3, 4, 4]],
'desc_bprop': [[1, 3, 4, 4]]}),
('TanhGrad', {
'block': G.TanhGrad(),
'desc_inputs': [[1, 3, 4, 4], [1, 3, 4, 4]],
'desc_bprop': [[1, 3, 4, 4]],
def __init__(self):
super().__init__()
self.matmul = P.MatMul(transpose_b=True)
self.gelu = P.Gelu()
def __init__(self):
super(VirtualDatasetNet, self).__init__()
self.virtual_dataset = _VirtualDataset()
self.matmul1 = P.MatMul()
self.matmul2 = P.MatMul()
self.gelu = P.Gelu()
def __init__(self, strategy1, strategy2):
super().__init__()
self.matmul = P.MatMul(transpose_b=True).set_strategy(strategy1)
self.gelu = P.Gelu().set_strategy(strategy2)
def __init__(self):
super(GELU, self).__init__()
self.gelu = P.Gelu()
def __init__(self, strategy0, strategy1, strategy2):
super().__init__()
self.fc_nobias = P.MatMul(transpose_b=True).shard(strategy0)
self.add = P.Add().shard(strategy1)
self.gelu = P.Gelu().shard(strategy2)
def __init__(self, 输入_接口, 输出_接口=2048, 丢弃率=0.1):
super(前向传播网络, self).__init__()
self.linear_1 = 全连接层(输入_接口, 输出_接口)
self.gelu = P.Gelu()
self.linear_2 = 全连接层(输出_接口, 输入_接口)
self.Dropout = nn.Dropout(1 - 丢弃率)
