From 226cbc05fd48a6d58b788af4e7cf52e902293383 Mon Sep 17 00:00:00 2001 From: Siddharth Date: Tue, 18 Jun 2019 22:17:43 +0530 Subject: [PATCH] corrected readme file --- TAN/README.md | 9 +++------ 1 file changed, 3 insertions(+), 6 deletions(-) diff --git a/TAN/README.md b/TAN/README.md index a436dd4..2c51549 100644 --- a/TAN/README.md +++ b/TAN/README.md @@ -1,4 +1,4 @@ -# networks.py +# networks.py ## `class LSTM_TAN(nn.Module):` ``` def __init__(self,version,embedding_dim, hidden_dim, vocab_size, n_targets,embedding_matrix,dropout = 0.5): @@ -24,7 +24,7 @@ __Args__ 2. target - a numpy array of shape [1xM] and dtype int, where M is the length of the target and each entry is the corresponding index of the word in the `embedding_matrix __Returns__ -1. target_scores - a torch float Tensor of shape [1xn_targets], where N is the number of dataset classes. This is the log likelihood probabilities of all the classes +1. target_scores - a torch float Tensor of shape [1xn_targets], where N is the number of dataset classes. This is the log likelihood probabilities of all the classes @@ -59,10 +59,7 @@ if dataset == 'EC': topic = "climate change is a real concern" elif dataset == 'FM' : topic = "feminist movement" - elif dataset == 'VCA': - topic = "vaccines cause autism" - elif dataset == 'VTI': - topic = "vaccines treat influenza" + ``` 2. version is one of ["lstm","tan-","tan+"]