Xuewen Shi
Xuewen Shi
Hi, I found some Chinese sentences (about 4000 sentences) in train.en file. for example  I'm not sure if these bugs will affect other parallel data. Thanks
https://github.com/jiasenlu/visDial.pytorch/blob/f1aa59704902bdbf4be8e507b30f55d87633debb/train/train_all.py#L77
https://github.com/jiasenlu/visDial.pytorch/blob/f1aa59704902bdbf4be8e507b30f55d87633debb/train/train_all.py#L252 ques_hidden1 = netE_d.init_hidden(opt.batchSize) why ques_hidden1 is one of the inputs of netE_g instead of ques_hidden2 which is generated from netE_g? Thank you very much.
https://github.com/jiasenlu/visDial.pytorch/blob/f1aa59704902bdbf4be8e507b30f55d87633debb/misc/dataLoader.py#L117 Hi, I found this code does nothing for ans_idx[i] is a float number while opt_ids is an int matrix. Did you mean "opt_ids = np.delete(opt_ids,self.ans_ids[index, i], 0)" instead of...