haoopan

The code about get the image features!

2

Hi Cadene, Thank you open the source.But I face the problem about download the feature files, It's very slow to download,about 100 k/s. So can you share the code about...

The question about KLD loss

1

Excuse me again! When I use the KLDivLoss as the criterion in ./block/models/criterions/kl_divergence.py rather than crossentropy, I get the loss is 0 and the acc is worse , as following...

Hi, It's a great job! In your semantic relationship encoding, you use one-hot encoding in semantic matrix. I wnat to use word embedding to encode the semantic relationships, so can...

The semantic_embedding and spatic_embedding types.

4

Hi, it is a great work for VQA. I did't download the datasets. So I want to konw the types of semantic_embedding and spatic_embedding, are they one-hot embedding or word...

请问你实现的动态卷积速度如何

3

感谢你的复现并开源，请问你有试过你写的动态卷积的速度吗，速度如何？谢谢！

I see your code in make tree and treelstm, find you use for loop to sperate each batch data. I think this may be reduce the training speed.

Hi, sorry to disturb you, I have a question when running your code on nlvr. When I remove the pre-trained model to train nlvr, the result is : Epoch 0:...