Conchylicultor

Sorry if I'm late. If the words are repeating, it may be because you don't train it long enough. It's usually the case at the beginning of the training. You...

The pre-trained model has been trained with a older​ version of tensorflow so it is possible that there is some compatibility issues with TF 1.1

For now the program does not support multi-GPU. Here is a example of multi GPU training if you need to implement it: https://github.com/tensorflow/tensorflow/blob/master/tensorflow/models/image/cifar10/cifar10_multi_gpu_train.py

There will be some major changes with tensoflow 1.0 for the [seq2seq API](https://github.com/tensorflow/tensorflow/tree/r1.0/tensorflow/contrib/seq2seq/python/ops). If multi-GPU is implemented someday, I don't think it would be pertinent to add that before tf...

You have to replace `` by the real path in the command `DEEPQA_WORKDIR= docker-compose -f`

You should not modify the deploy.yml file

From what I remember changing the max length will change the vocabulary size used. So the program will crash because the vocabulary embedding shape does not match. I think it...

Try to set the `softmaxSamples` parameter: `--softmaxSamples 512`. That may helps. Also I'll say that the vocabulary size is probably too big as described here https://github.com/Conchylicultor/DeepQA/issues/29#issuecomment-267771058.

My suggestion was to filter the words which appears rarely on the dataset and replace them by the unknown token. I don't really see any other trivial solution.

I'm not sure I understand the question. Each sentence is directly split into token and then each token is added to a dictionary self.word2id. For now, no frequency count is...