Vadim Kantorov

ITER_SIZE different for R50 and WSR50 (pascal voc detection)

1

Why would there be such a difference? Thanks!

How important is dropout in DAN?

1

Droput prob here is quite high - 0.5, and it is not discussed in the paper...

What is meaning of MEAN_LOSS = False | True

5

What's effective loss scaling? Does it sum or mean over classes? over batch size? How does it interact with distributed training? Is there anywhere scaling over the world size?

Number of epochs for VOC07 / WSDDN difference for R50 and WSR50

2

WSR50 config has 160 epochs: https://github.com/shenyunhang/DRN-WSOD-pytorch/blob/DRN-WSOD/projects/WSL/configs/PascalVOC-Detection/wsddn_WSR_50_DC5_1x.yaml R50 config has 28 epochs: https://github.com/shenyunhang/DRN-WSOD-pytorch/blob/DRN-WSOD/projects/WSL/configs/PascalVOC-Detection/wsddn_R_50_DC5_1x.yaml (this is true for WSR and R configs in gneral) Why is this the case? Thanks!

_FieldSkipper problem

1

I am trying to parse a Caffe [model](https://github.com/BVLC/caffe/blob/master/src/caffe/proto/caffe.proto). I am getting an error related to _FieldSkipper: ``` /sequoia/data2/kantorov/wigwam_sequoia_gpu101_105/.wigwam/prefix/bin/luajit: ...01_105/.wigwam/prefix/share/lua/5.1/protobuf/decoder.lua:333: attempt to call a nil value stack traceback: ...01_105/.wigwam/prefix/share/lua/5.1/protobuf/decoder.lua:333: in function...

Question about implementing a pairwise L2 distances function

1

Hi, I'm trying to implement an L2-distances module. The code below prints: ``` function: 0x41cd5570 3.999683343136 ``` which means the gradient check doesn't pass. What am I missing? ``` lua...

@danielbear How could one ask for access for Playroom and Primitives datasets? Pixelwise labels are very useful for object-centric research. Thanks!

Inplace fused (leaky-)relu+dropout for memory savings

14

### 🚀 The feature, motivation and pitch It's possible that such an inplace fusion is already possible with dynamo, but if not, it's quite good to have for saving memory....

### 📚 The doc issue https://pytorch.org/docs/stable/generated/torch.nn.functional.scaled_dot_product_attention.html?highlight=scaled_dot_product_attention#torch.nn.functional.scaled_dot_product_attention: The code example is rather not ordinary and contains a `return` statement outside of any function definition: is `not attn_mask` correct? should it not...

Hi! In https://github.com/MarcoForte/closed-form-matting/blob/master/closed_form_matting/closed_form_matting.py#L82 `cv2.dilate` is applied to the mask, while in `matting.tar.gz` is applied `imerode` If I understand correctlu, `imerode` is the opposite of `imdilate` (https://docs.opencv.org/3.4/db/df6/tutorial_erosion_dilatation.html). Any hints on why...