# Extract feature maps from intermediate layers without modifying forward() **URL:** \ **Category:** vision\ **Created:** [March 27, 2017, 7:20am UTC](https://discuss.pytorch.org/t/extract-feature-maps-from-intermediate-layers-without-modifying-forward/1390 "2017-03-27T07:20:44Z")\ **Posts on this page:** 7\ **Page:** 1
**Author:** ![jsalbert](https://discuss.pytorch.org/letter_avatar_proxy/v4/letter/j/8491ac/32.png) [@jsalbert](https://discuss.pytorch.org/u/jsalbert)\ **Post date:** [March 27, 2017, 7:20am UTC](https://discuss.pytorch.org/t/extract-feature-maps-from-intermediate-layers-without-modifying-forward/1390/1 "2017-03-27T07:20:45Z")
Hi, I am interested in obtaining features from the intermediate layers of my model, but without modifying the forward() method of the model, as it is already trained. And also I don’t want to split it, because I am interested in getting the prediction, and the features from other upper layers as well in the same forward pass. I have read about the register\_forward\_hook, but I haven’t found any example on how to use it. I have this: ``` def get_features_hook(self, input, output): print output.data.cpu().numpy().shape features = output.data.cpu().numpy() model.features.module[37].register_forward_hook(get_features_hook) model.forward(im_tensor) ``` Is there any way to extract that features value? Thanks ---
**Author:** ![smth](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/smth/32/13_2.png) [@smth](https://discuss.pytorch.org/u/smth)\ **Post date:** [March 27, 2017, 2:06pm UTC](https://discuss.pytorch.org/t/extract-feature-maps-from-intermediate-layers-without-modifying-forward/1390/2 "2017-03-27T14:06:54Z")
There are many examples in this thread: > [@Why cant I see .grad of an intermediate variable?](https://discuss.pytorch.org/t/why-cant-i-see-grad-of-an-intermediate-variable/94): > > Hi! I am loving this framework… Since Im a noob, I am probably not getting something, but I am wondering why I cant get the gradient of an intermediate variable with .grad? Here is an example of what I mean: xx = Variable(torch.randn(1,1), requires\_grad = True) yy = 3\*xx zz = yy\*\*2 zz.backward() xx.grad # This is ok yy.grad # This gives 0! zz.grad # This should give 1! So I get the correct result for xx.grad, but why does yy.grad show 0, as does zz.grad? How can I get the yy.grad value in t… Also a search will give you a few more examples: [https://discuss.pytorch.org/search?q=register\_hook](https://discuss.pytorch.org/search?q=register_hook) ---
**Author:** ![bhattad2](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/bhattad2/32/3346_2.png) [@bhattad2](https://discuss.pytorch.org/u/bhattad2)\ **Post date:** [June 27, 2017, 10:25pm UTC](https://discuss.pytorch.org/t/extract-feature-maps-from-intermediate-layers-without-modifying-forward/1390/4 "2017-06-27T22:25:30Z")
Hi, I am trying to extract feature outputs of the intermediate layers of pre-trained VGG 16 architecture and concatenate them. The in-built models in pytorch doesn’t have names for all its layers for VGG architecture. Therefore, I am unable to use register\_forward\_hook. Is there any other alternate way? I am trying to use something like below but I am not sure if gradients will be accumulated at the intermediate layers when I do back propagation as it is building up two computation graphs during instantiation of my model. I have something like this ``` def __init__ (self): super(ModifiedVGG, self). __init__ () k = [29, 22] model = models.vgg16(pretrained=True) self.layer1 = nn.Sequential(*list(model.features.children())[:k[0]]) self.layer2 = nn.Sequential(*list(model.features.children())[:k[1]]) self.conv1 = nn.Conv2d(512, 128, 3, padding=1) self.upsample1 = nn.UpsamplingBilinear2d(scale_factor=2) def forward(self, x): conv_4_3 = self.layer2(x) conv_4_3 = F.relu(self.conv1(conv_4_3)) conv_5_3 = self.layer1(x) conv_5_3 = F.relu(self.conv1(conv_5_3)) conv_5_3 = self.upsample1(conv_5_3) concat_features = torch.cat([conv_4_3, conv_5_3], 1) return concat_features ``` Please let me know if this would work or not. Thank you! ---
**Author:** ![smth](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/smth/32/13_2.png) [@smth](https://discuss.pytorch.org/u/smth)\ **Post date:** [June 27, 2017, 10:37pm UTC](https://discuss.pytorch.org/t/extract-feature-maps-from-intermediate-layers-without-modifying-forward/1390/5 "2017-06-27T22:37:49Z")
the simplest solution is to: - copy over the vgg file - change the name of the class - modify the class (for example remove or replace layers as you see fit) - use the function [load\_state\_dict](http://pytorch.org/docs/master/nn.html?highlight=load_state_dict#torch.nn.Module.load_state_dict) to load the original VGG weights dict into this modified class. This is less error prone, and it’s easily modifiable. The vgg weights state dict is available in this URL: [https://github.com/pytorch/vision/blob/master/torchvision/models/vgg.py#L12-L21](https://github.com/pytorch/vision/blob/master/torchvision/models/vgg.py#L12-L21) ---
**Author:** ![bhattad2](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/bhattad2/32/3346_2.png) [@bhattad2](https://discuss.pytorch.org/u/bhattad2)\ **Post date:** [June 27, 2017, 10:56pm UTC](https://discuss.pytorch.org/t/extract-feature-maps-from-intermediate-layers-without-modifying-forward/1390/6 "2017-06-27T22:56:27Z")
Thank you. This would be the least error prone of all. Also, when I use register\_forward\_hook, do I need to worry about backward\_hook or it will be taken care off automatically? Some more examples on register\_hook would be appreciated. Thank you. ---
**Author:** ![ne-bo](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/ne-bo/32/20651_2.png) [@ne-bo](https://discuss.pytorch.org/u/ne-bo)\ **Post date:** [January 22, 2018, 12:08pm UTC](https://discuss.pytorch.org/t/extract-feature-maps-from-intermediate-layers-without-modifying-forward/1390/7 "2018-01-22T12:08:11Z")
Thanks a lot this code works for me! ---
**Author:** ![ssh983](https://discuss.pytorch.org/letter_avatar_proxy/v4/letter/s/b5a626/32.png) [@ssh983](https://discuss.pytorch.org/u/ssh983)\ **Post date:** [March 2, 2018, 5:01pm UTC](https://discuss.pytorch.org/t/extract-feature-maps-from-intermediate-layers-without-modifying-forward/1390/8 "2018-03-02T17:01:20Z")
Hey @smth, I have looked over a lot, but couldnt find an example of this method. Could you please point to any of it? I am just starting pyTorch, and couldnt find much tuts on how about using this function for transfer learning.