indrnn-pytorch
indrnn-pytorch copied to clipboard
StefOe
pytorch implementation of Independently Recurrent Neural Networks https://arxiv.org/abs/1803.04831
Hi, Thank you for the code! I run the code without any modification and I got 97% acc for MNIST and 52% acc for pMNIST. Did you manage to reproduce...
Hi, I want to use indRNN with batches of examples of different size, but I see this in the repo: ``` if self.bidirectional: x_T = torch.cat([x, x.flip(-1)], 1) ``` This...
Do you plan to test this model on the action recognition dataset? The performance of my implementation is very bad.
https://github.com/StefOe/indrnn-pytorch/blob/master/indrnn.py class IndRNN(nn.Module) describes that input can be the PackedSequence but it is not. For example, line 232: x.size(batch_index) makes error AttributedError: 'PackedSequence' object has no attribute 'size'
Dropout
The paper mentions > "Dropout [9] was applied after each IndRNN layer with a dropping probability of 0.25 and 0.1 for CS and CV settings, respectively." and > "Dropout [9]...
I think the check_bounds has done the same operation.
@StefOe hello StefOe， i just wonder if it is possible to add a new feature :bidirectional
Hi, I have uploaded my code of IndRNN to GitHub as well as some running examples using it. https://github.com/Sunnydreamrain/IndRNN_Theano_Lasagne Since there are a few items in your to-do list, I...
