pytorch网络转libtorch常见问题

目录

---

一、RuntimeError: all inputs of range must be ints, found Tensor in argument 0:

二、RuntimeError: Sliced expression not yet supported for subscripted assignment. File a bug if you want this:

三、Tried to access nonexistent attribute or method 'len' of type 'torch.torch.nn.modules.container.ModuleList'. Did you forget to initialize an attribute in init()?

四、Expected integer literal for index

---

一、RuntimeError: all inputs of range must be ints, found Tensor in argument 0:

问题
参数类型不正确，函数的默认参数是tensor

解决措施
函数传入参数不是tensor需要注明类型
我的问题是传入参数npoint是一个int类型，没有注明会报错，更改如下：
由

def test(npoint):
  ...

更改为

def test(npoint: int):
  ...

---

二、RuntimeError: Sliced expression not yet supported for subscripted assignment. File a bug if you want this:

问题
不支持赋值给切片表达式

解决措施
根据自己需求，进行修改，可利用循环替代

我将view_shape[1:] = [1] * (len(view_shape) - 1)更改为

    for i in range(1, len(view_shape)):
        view_shape[i] = 1

---

三、Tried to access nonexistent attribute or method 'len' of type 'torch.torch.nn.modules.container.ModuleList'. Did you forget to initialize an attribute in init()?

问题
forward函数中好像不支持len(nn.ModuleList())和下标访问

解决措施
如果是一个ModuleList()可以用enumerate函数，多个同维度的可以用zip函数

我这里有两个ModuleList()，所以采用zip函数，更改如下：
由

   for i, conv in enumerate(self.mlp_convs):
      bn = self.mlp_bns[i]
      new_points = F.relu(bn(conv(new_points)))

更改为

    for conv, bn in zip(self.mlp_convs, self.mlp_bns):
        new_points = F.relu(bn(conv(new_points)))

ref: https://github.com/pytorch/pytorch/issues/16123

---

四、Expected integer literal for index

问题和解决方法类似第三个

标签：常见问题,torch,pytorch,len,libtorch,shape,new,ModuleList,view
来源： https://www.cnblogs.com/xiaxuexiaoab/p/15555066.html