1 系统环境
硬件环境(Ascend/GPU/CPU): CPU
MindSpore版本: 1.7.0
执行模式（PyNative/ Graph）: Graph
Python版本: Python=3.9.15
操作系统平台: 不限
2 报错信息
2.1 问题描述
静态图模式下的construct函数中对SequentialCell通过下标访问元素出错，“SequentialCell[-1]” 语法不支持。
2.2 报错信息
RuntimeError: The 'getitem' operation does not support the type [Func, Int64].
2.3 脚本代码
from mindspore import nn
from mindspore.nn import Cell
from mindspore import context
import mindspore
from mindspore.ops import operations as op
import numpy as np
from mindspore import Tensor
context.set_context(mode=context.GRAPH_MODE)
class Network(nn.Cell):
def __init__(self):
super().__init__()
self.flatten = nn.Flatten()
self.dense_relu_sequential = nn.SequentialCell(
nn.Dense(28*28, 512),
nn.ReLU(),
nn.Dense(512, 512),
nn.ReLU(),
nn.Dense(512, 10)
)
self.print = op.Print()
def construct(self, x):
x = self.flatten(x)
logits = self.dense_relu_sequential[-1](x)
return logits
X = Tensor(np.ones((1, 16, 32), np.float32))
model = Network()
logits = model(X)
3 根因分析
静态图模式不支持SequentialCell的getitem操作。如果需要执行getitem操作，建议使用nn.CellList。
4 解决方案
将SequentialCell换成CellList即可;
CellList可以像普通Python列表一样使用，其包含的Cell均已初始化。
from mindspore import nn
from mindspore.nn import Cell
from mindspore import context
import mindspore
from mindspore.ops import operations as op
import numpy as np
from mindspore import Tensor
context.set_context(mode=context.GRAPH_MODE)
class Network(nn.Cell):
def __init__(self):
super().__init__()
self.flatten = nn.Flatten()
self.dense_relu_sequential = nn.CellList(
[nn.Dense(28*28, 512),
nn.ReLU(),
nn.Dense(512, 512),
nn.ReLU(),
nn.Dense(512, 10)]
)
self.print = op.Print()
def construct(self, x):
x = self.flatten(x)
logits = self.dense_relu_sequential[-1](x)
return logits
X = Tensor(np.ones((1, 16, 32), np.float32))
model = Network()
logits = model(X)
执行无报错
