deb

model/crf.py

@@ -35,11 +35,10 @@ def rand_init(self):
 
     def forward(self, feats):
         """
-        inputs: 
-            - **feats** (batch_size, seq_len, hidden_dim) : input score from previous layers
-
-        outputs:
-            - **crf_score** (batch_size, seq_len, tag_size, tag_size) : output from crf layer
+        args: 
+            feats (batch_size, seq_len, hidden_dim) : input score from previous layers
+        return:
+            Tensor (batch_size, seq_len, tag_size, tag_size) : output from crf layer
         """
         return self.hidden2tag(feats).view(-1, self.tagset_size, self.tagset_size)
 
@@ -68,11 +67,10 @@ def rand_init(self):
 
     def forward(self, feats):
         """
-        inputs: 
-            - **feats** (batch_size, seq_len, hidden_dim) : input score from previous layers
-
-        outputs:
-            - **crf_score** (batch_size, seq_len, tag_size, tag_size) : output from crf layer
+        args: 
+            feats (batch_size, seq_len, hidden_dim) : input score from previous layers
+        return:
+            Tensor (batch_size, seq_len, tag_size, tag_size) : output from crf layer
 
         """
 
@@ -98,14 +96,12 @@ def __init__(self, tagset_size, if_cuda):
     def repack_vb(self, feature, target, mask):
         """packer for viterbi loss
 
-        inputs: 
+        args: 
             - **feature** (Seq_len, Batch_size): input feature
             - **target** (Seq_len, Batch_size): output target
             - **mask** (Seq_len, Batch_size): padding mask
         outputs:
-            - **feature** (Seq_len, Batch_size) : input feature
-            - **target** (Seq_len, Batch_size) : output target
-            - **mask** (Seq_len, Batch_size) : padding mask
+            - **feature**, **target**, **mask**
         """
 
         if self.if_cuda:
@@ -121,7 +117,7 @@ def repack_vb(self, feature, target, mask):
     def repack_gd(self, feature, target, current):
         """packer for greedy loss
 
-        argss: 
+        args: 
             feature: input feature, of size Seq_len * Batch_size
             target: output target, of size Seq_len * Batch_size
             current: current state, of size Seq_len * Batch_size