finish train.py

MainPaper/train.py

@@ -11,6 +11,8 @@
 from util import *
 import matplotlib.pyplot as plt
 
+import time
+
 
 DATA_SUMMARY_CSV_PATH = '../data_summary.csv'
 DATA_SUMMARY_HEADER =  {"person":"person_name", "video":"video_name", "frame":"frame_number", "pspi":"pspi_score", "image":"image_path"}
@@ -49,12 +51,12 @@ def print_opts(opts):
     'image_scale_to_before_crop': 320,
     'image_size': 160,
     'number_output': 1,
-    'image_sample_size': 1,
-    'batch_size': 100,  
+    'image_sample_size': 5,
+    'batch_size': 50,  
     'drop_out': 0,
     'fc2_size': 200,
     'learning_rate': 0.001,
-    'epoch': 1
+    'epoch': 10
 }
 ARGS.update(args_dict)
 #endregion
@@ -153,7 +155,7 @@ def train(train_data_path, ops):
         title = f"Epoch={epoch}, Train loss of each batch"
         x_label = "Batch number"
         y_label = "Loss"
-        save_path = os.path.join("./result", f"Epoch_{epoch}_loss.png")
+        save_path = os.path.join("./result", f"Epoch_{epoch+1}_loss.png")
         draw_line_chart(x_list, train_loss_for_each_batch_list, title, x_label, y_label, save_path)
 
         avg_loss_for_each_epoch = np.array(train_loss_for_each_batch_list).mean()
@@ -172,7 +174,10 @@ def train(train_data_path, ops):
 
 
 
-def evaluation(net, test_data_loader):
+def evaluation(net, test_data_path, ops):
+    test_dataset = MyDataset(test_data_path, ops)
+    test_data_loader = torch.utils.data.DataLoader(test_dataset, batch_size=ops.batch_size, shuffle=True)
+
     mse_loss = nn.MSELoss()
 
     total_number_batch = 0
@@ -204,8 +209,13 @@ def main():
     create_data_csv(DATA_CSV_PATH, DATA_SUMMARY_CSV_PATH, ARGS)
     split_dataset(DATA_CSV_PATH, TRAIN_DATA_CSV_PATH, TEST_DATA_CSV_PATH, RANDOM_SEED, TRAIN_FRACTION)
 
+    start = time.time()
     net = train(TRAIN_DATA_CSV_PATH, ARGS)
-    evaluation(net, TEST_DATA_CSV_PATH)
+    end = time.time()
+    print(f"Runtime of the program is {end - start}")
+
+
+    evaluation(net, TEST_DATA_CSV_PATH, ARGS)