3 Commits

Author	SHA1	Message	Date
James Reed	10d95bd0f0	Remove batch_size parameter from attention and LSTMWithAttention interfaces ... Summary: Reshape based on tensor shapes in the graph rather than based on a passed-in batch_size parameter Reviewed By: urikz Differential Revision: D4702086 fbshipit-source-id: c4c1d8425cd36c1e86695918eaba2667c27e9601	2017-03-16 11:47:52 -07:00
James Reed	8de1db9eb6	Implement recurrent attention in C2 ... Summary: Super rough implementation of recurrent attention. Planning to factor out the common code between the two functions as well as train and eval. I want to get this out and get eyes on it sooner rather than later Differential Revision: D4647837 fbshipit-source-id: 54bc4e8ed0df6f04c86c425926decbe89f73b068	2017-03-08 11:21:28 -08:00
Yury Zemlyanskiy	4a53ab3cb6	LSTMWithAttention implementation in Caffe2 ... Summary: Implementation of ##LSTMWithAttention## Still TBD: 1. There are problems with back propagation, because gradient is not implemented for ops with broadcasting 2. I need to make initial_recurrent_state to be of shape [dim] rather than [1, batch_size, dim], so one doesn't need to provide batch_size to LSTMWithAttention Differential Revision: D4298735 fbshipit-source-id: 8903fcff4d6a66647ee6d45a6ef28803fc3091e5	2017-02-23 04:08:34 -08:00