rename to re_parameterize() for re-parameterization nets

ppcls/arch/backbone/legendary_models/pp_lcnet_v2.py

@@ -216,7 +216,7 @@ class RepDepthwiseSeparable(TheseusLayer):
             x = x + input_x
         return x
 
-    def rep(self):
+    def re_parameterize(self):
         if self.use_rep:
             self.is_repped = True
             kernel, bias = self._get_equivalent_kernel_bias()

ppcls/arch/backbone/model_zoo/repvgg.py

@@ -172,7 +172,7 @@ class RepVGGBlock(nn.Layer):
         return self.nonlinearity(
             self.se(self.rbr_dense(inputs) + self.rbr_1x1(inputs) + id_out))
 
-    def rep(self):
+    def re_parameterize(self):
         if not hasattr(self, 'rbr_reparam'):
             self.rbr_reparam = nn.Conv2D(
                 in_channels=self.in_channels,

ppcls/arch/slim/quant.py

@@ -49,10 +49,10 @@ def quantize_model(config, model, mode="train"):
         if mode in ["infer", "export"]:
             QUANT_CONFIG['activation_preprocess_type'] = None
 
-        # for rep nets, convert to reparameterized model first
+        # for re-parameterization nets, convert to reparameterized model first
         for layer in model.sublayers():
-            if hasattr(layer, "rep"):
-                layer.rep()
+            if hasattr(layer, "re_parameterize"):
+                layer.re_parameterize()
 
         model.quanter = QAT(config=QUANT_CONFIG)
         model.quanter.quantize(model)

ppcls/engine/engine.py

@@ -560,10 +560,11 @@ class Engine(object):
 
         model.eval()
 
-        # for rep nets
+        # for re-parameterization nets
         for layer in self.model.sublayers():
-            if hasattr(layer, "rep") and not getattr(layer, "is_repped"):
-                layer.rep()
+            if hasattr(layer, "re_parameterize") and not getattr(layer,
+                                                                 "is_repped"):
+                layer.re_parameterize()
 
         save_path = os.path.join(self.config["Global"]["save_inference_dir"],
                                  "inference")