How to obtain results from certain layer from a pretrained model which isn't written using Sequential()?

Hi, I tried to return the output from an intermediate layer of a pretrained model. Have searched many times but most of the examples are for model that’s built with Sequential() method or from model Zoo. My model is a customized model (no Sequential is used here) and when I called model._modules, it only returns an OrderDict, but the order of operation is not the same as during forward(). So directly calling module(x) doesn’t work! My question is how to do this for model that’s not neither Sequential() nor VGG (models that are already in model zoo)?

Here’s my code but it doesn’t work:

class SelectiveLayer(nn.Module):
	def __init__(self, to_select, model):
		super(SelectiveLayer, self).__init__()
		for key, module in model._modules.items():
			self.add_module(key, module)
		self._to_select = to_select

	def forward(self, x):
		list = []
		for name, module in self._modules.items():
			print(name, module, x.shape)
			x = module(x)
			if name in self._to_select:
		return list

class PerceptualLoss(nn.Module):
 	def __init__(self, pretrained_model, params):
 		super(PerceptualLoss, self).__init__()
 		self.pretrained_model = pretrained_model
 		self.features = SelectiveLayer("conv5", pretrained_model)

 	def forward(self, x):
 		x = self.features(x)
 		return x

Hi, you can use hooks. Small tutorial

Updates here. I used hooks given the information from Also thanks @JuanFMontesinos for the information.