Update modeling_super_linear.py

modeling_super_linear.py

@@ -379,9 +379,9 @@ class superLinear(nn.Module):
 
         self.layer_dict = {'DLinear': DLinear, 'Linear': Linear, 'NLinear': NLinear, 'RLinear': RLinear}
         
-        path = configs.linear_checkpoints_path + configs.linear_checkpoints_dir 
-        dirs = os.listdir(path)
-        checkpoints_paths = [path + "/" + d + "/" + "checkpoint.pth" for d in dirs]
+        # path = configs.linear_checkpoints_path + configs.linear_checkpoints_dir 
+        # dirs = os.listdir(path)
+        # checkpoints_paths = [path + "/" + d + "/" + "checkpoint.pth" for d in dirs]
 
         if self.freq_experts == "all":
             self.freq_experts = []
@@ -425,11 +425,11 @@ class superLinear(nn.Module):
 
         if configs.misc_moe>0:
             if configs.misc_moe == 1:
-                #print("Creating misc expert")
+                print("Creating misc expert")
                 self.experts["misc"] = self.layer_dict[self.layer_type](self.seq_len, self.pred_len)
             else:
                 for i in range(configs.misc_moe):
-                    #print(f"Creating misc expert {i}")
+                    print(f"Creating misc expert {i}")
                     self.experts["misc_"+str(i)] = self.layer_dict[self.layer_type](self.seq_len, self.pred_len)
 
 
@@ -437,18 +437,18 @@ class superLinear(nn.Module):
         self.moe = SparseNoisyMoE(configs, experts=self.experts.values())
         self.dropout = nn.Dropout(configs.dropout)
 
-        if configs.load_weights:
-            print(f"Loading weights from {path}")
-            path = configs.load_weights_path + "" + configs.load_weights_dir + "/" + "checkpoint.pth"
-            if os.path.exists(path):
-                checkpoint = torch.load(path)
-                print(len(self.experts.keys()))
-                print(self.experts.keys())
-                print(self.state_dict().keys())
-                print(checkpoint.keys())
-                self.load_state_dict(checkpoint)
-            else:   
-                print(f"Path {path} does not exist. Skipping loading weights.")
+        # if configs.load_weights:
+        #     print(f"Loading weights from {path}")
+        #     path = configs.load_weights_path + "" + configs.load_weights_dir + "/" + "checkpoint.pth"
+        #     if os.path.exists(path):
+        #         checkpoint = torch.load(path)
+        #         print(len(self.experts.keys()))
+        #         print(self.experts.keys())
+        #         print(self.state_dict().keys())
+        #         print(checkpoint.keys())
+        #         self.load_state_dict(checkpoint)
+        #     else:   
+        #         print(f"Path {path} does not exist. Skipping loading weights.")
 
 
     def map_to_cycle(self, freq):