我有一个模块列表，我想使用另一个索引列表将其编入索引

import torch
import torchvision.transforms as T

transforms = torch.nn.ModuleList([T.ColorJitter(), T.ColorJitter()])
order = [1,0]

字符串
I cannot do transforms[order[0]]。但是，我可以迭代ModuleList：for t in transforms:和for i,t in enumerate(transforms):工作
如何通过迭代或枚举有效地索引到ModuleList中？
我尝试了以下方法，但它们不起作用

# Permute/Change ordering of the ModuleList using a ModuleDict and then iterate the ModuleDict
permuted_transforms = torch.nn.ModuleDict({order[i]:t for i,t in enumerate(transforms)})

给出FrontendError: Cannot instantiate class 'ModuleDict' in a script function
*

# Permute/Change ordering of the ModuleList using torch.take
permuted_transforms = torch.take(self.transforms, order)

但是torch.take只对torch.tensors有效，而对ModuleLists无效
*

# Permute/Change ordering of the ModuleList using map
permuted_transforms = map(self.transforms.__getitem__, order)

型
*

# Permute/Change ordering of the ModuleList using sorted
permuted_transforms = sorted(self.transforms, key=order.__getitem__)

型
*

# Have 2 for loops work but is extremely ineffecient
for o in order:
  for i,t in enumerate(transforms):
    if i==o: apply(t)

型