add embedding and fix eval when distillation (#3112)

configs/rec/ch_ppocr_v2.1/rec_chinese_lite_train_distillation_v2.1.yml

@@ -52,9 +52,10 @@ Architecture:
       Neck:
         name: SequenceEncoder
         encoder_type: rnn
-        hidden_size: 48
+        hidden_size: 64
       Head:
         name: CTCHead
+        mid_channels: 96
         fc_decay: 0.00001
     Teacher:
       pretrained:
@@ -71,9 +72,10 @@ Architecture:
       Neck:
         name: SequenceEncoder
         encoder_type: rnn
-        hidden_size: 48
+        hidden_size: 64
       Head:
         name: CTCHead
+        mid_channels: 96
         fc_decay: 0.00001
   
 

ppocr/modeling/heads/rec_ctc_head.py

@@ -33,19 +33,47 @@ def get_para_bias_attr(l2_decay, k):
 
 
 class CTCHead(nn.Layer):
-    def __init__(self, in_channels, out_channels, fc_decay=0.0004, **kwargs):
+    def __init__(self,
+                 in_channels,
+                 out_channels,
+                 fc_decay=0.0004,
+                 mid_channels=None,
+                 **kwargs):
         super(CTCHead, self).__init__()
-        weight_attr, bias_attr = get_para_bias_attr(
-            l2_decay=fc_decay, k=in_channels)
-        self.fc = nn.Linear(
-            in_channels,
-            out_channels,
-            weight_attr=weight_attr,
-            bias_attr=bias_attr)
+        if mid_channels is None:
+            weight_attr, bias_attr = get_para_bias_attr(
+                l2_decay=fc_decay, k=in_channels)
+            self.fc = nn.Linear(
+                in_channels,
+                out_channels,
+                weight_attr=weight_attr,
+                bias_attr=bias_attr)
+        else:
+            weight_attr1, bias_attr1 = get_para_bias_attr(
+                l2_decay=fc_decay, k=in_channels)
+            self.fc1 = nn.Linear(
+                in_channels,
+                mid_channels,
+                weight_attr=weight_attr1,
+                bias_attr=bias_attr1)
+
+            weight_attr2, bias_attr2 = get_para_bias_attr(
+                l2_decay=fc_decay, k=mid_channels)
+            self.fc2 = nn.Linear(
+                mid_channels,
+                out_channels,
+                weight_attr=weight_attr2,
+                bias_attr=bias_attr2)
         self.out_channels = out_channels
+        self.mid_channels = mid_channels
 
     def forward(self, x, labels=None):
-        predicts = self.fc(x)
+        if self.mid_channels is None:
+            predicts = self.fc(x)
+        else:
+            predicts = self.fc1(x)
+            predicts = self.fc2(predicts)
+
         if not self.training:
             predicts = F.softmax(predicts, axis=2)
         return predicts

tools/eval.py

@@ -44,8 +44,15 @@ def main():
     # build model
     # for rec algorithm
     if hasattr(post_process_class, 'character'):
-        config['Architecture']["Head"]['out_channels'] = len(
-            getattr(post_process_class, 'character'))
+        char_num = len(getattr(post_process_class, 'character'))
+        if config['Architecture']["algorithm"] in ["Distillation",
+                                                   ]:  # distillation model
+            for key in config['Architecture']["Models"]:
+                config['Architecture']["Models"][key]["Head"][
+                    'out_channels'] = char_num
+        else:  # base rec model
+            config['Architecture']["Head"]['out_channels'] = char_num
+
     model = build_model(config['Architecture'])
     use_srn = config['Architecture']['algorithm'] == "SRN"