Upload folder using huggingface_hub

csv/check.py

@@ -0,0 +1,545 @@
+# check evaluated number of images
+# check score in list, score 0, 0.5, 1
+# check total score_alignment, score_quality
+
+import csv
+import os
+import ast
+
+csv_file = 'your_file.csv'
+folder_path = 'your_folder'
+
+
+def check(csv_file, image_folder, s1, s2, s3, s4, s5):
+    print(f"###################{csv_file}#######################")
+    # 1. Count CSV rows (excluding header)
+    with open(csv_file, newline='') as f:
+        reader = list(csv.DictReader(f))
+        num_rows = len(reader)
+
+    # 2. Count files in folder
+    num_files = len([name for name in os.listdir(image_folder) if name.endswith('.png') or name.endswith('.jpg')])
+
+    print(f"CSV rows: {num_rows}, Files in folder: {num_files}")
+    if num_rows != num_files:
+        print("Row and image count do not match!")
+    else:
+        print("Row and image count match.")
+
+    # 3. Check 's1', 's2', 's3' columns and calculate averages
+    valid_values = {0, 0.5, 1}
+    s1_averages, s2_averages, s3_averages = [], [], []
+    invalid_rows = []
+    
+    s4_values, s5_values = [], []
+    invalid_s4_s5_rows = []
+
+    for idx, row in enumerate(reader):
+        valid_row = True
+        row_avgs = []
+        for col in [s1, s2, s3]:
+     
+            vals = ast.literal_eval(row[col])  # Parse the list
+            if not isinstance(vals, list):
+                # raise ValueError  # not list
+                valid_row = False
+            if not all(v in valid_values for v in vals):
+                valid_row = False # not 0, 0.5, or 1
+            else:
+                avg = sum(vals) / len(vals) if vals else 0
+                row_avgs.append(avg)  
+        if not valid_row:
+            invalid_rows.append(idx+2)  # +2 for header and 0-indexing
+        if valid_row:
+            s1_averages.append(row_avgs[0])
+            s2_averages.append(row_avgs[1])
+            s3_averages.append(row_avgs[2])
+            
+        # Check s4 and s5
+        s4_ok = float(row[s4])<=1.0 and float(row[s4])>=0.0
+        s5_ok = float(row[s5])<=1.0 and float(row[s5])>=0.0
+        if s4_ok and s5_ok:
+            s4_values.append(float(row[s4]))
+            s5_values.append(float(row[s5]))
+        else:
+            invalid_s4_s5_rows.append(idx+2)
+            print(type(float(row[s4])))
+            
+
+    if invalid_rows:
+        print(f"Invalid rows in 's1', 's2', or 's3': {invalid_rows}")
+    else:
+        print("All rows in 's1', 's2', and 's3' are valid.")
+
+    if invalid_s4_s5_rows:
+        print(f"Invalid rows in 's4' or 's5': {invalid_s4_s5_rows}")
+    else:
+        print("All rows in 's4' and 's5' are valid.")
+        
+    overall_s1_avg = sum(s1_averages) / len(s1_averages) if s1_averages else 0
+    overall_s2_avg = sum(s2_averages) / len(s2_averages) if s2_averages else 0
+    overall_s3_avg = sum(s3_averages) / len(s3_averages) if s3_averages else 0
+
+    print(f"Average for s1: {overall_s1_avg:.3f}")
+    print(f"Average for s2: {overall_s2_avg:.3f}")
+    print(f"Average for s3: {overall_s3_avg:.3f}")
+    
+    overall_s4_avg = sum(s4_values) / len(s4_values) if s4_values else 0
+    overall_s5_avg = sum(s5_values) / len(s5_values) if s5_values else 0
+    print(f"Average for s4: {overall_s4_avg:.3f}")
+    print(f"Average for s5: {overall_s5_avg:.3f}")
+    
+    return overall_s1_avg, overall_s2_avg, overall_s3_avg, overall_s4_avg, overall_s5_avg
+
+
+def check2(csv_file, image_folder, s1, s2, s3, s4, s5, s6):
+    print(f"###################{csv_file}#######################")
+    # 1. Count CSV rows (excluding header)
+    with open(csv_file, newline='') as f:
+        reader = list(csv.DictReader(f))
+        num_rows = len(reader)
+
+    # 2. Count files in folder
+    num_files = len([name for name in os.listdir(image_folder) if name.endswith('.png') or name.endswith('.jpg')])
+
+    print(f"CSV rows: {num_rows}, Files in folder: {num_files}")
+    if num_rows != num_files:
+        print("Row and image count do not match!")
+    else:
+        print("Row and image count match.")
+
+    # 3. Check 's1', 's2', 's3' columns and calculate averages
+    valid_values = {0, 0.5, 1}
+    s1_averages, s2_averages, s3_averages = [], [], []
+    invalid_rows = []
+    
+    s4_values, s5_values, s6_values = [], [], []
+    invalid_s4_s5_s6_rows = []
+    
+    s45_values = []
+    w1 = 0.7
+    w2 = 0.3
+
+    for idx, row in enumerate(reader):
+        valid_row = True
+        row_avgs = []
+        for col in [s1, s2, s3]:
+     
+            vals = ast.literal_eval(row[col])  # Parse the list
+            if not isinstance(vals, list):
+                # raise ValueError  # not list
+                valid_row = False
+            if not all(v in valid_values for v in vals):
+                valid_row = False # not 0, 0.5, or 1
+                print(vals)
+                print("HHHHHHH2")
+            else:
+                avg = sum(vals) / len(vals) if vals else 0
+                row_avgs.append(avg)  
+        if not valid_row:
+            invalid_rows.append(idx+2)  # +2 for header and 0-indexing
+        if valid_row:
+            s1_averages.append(row_avgs[0])
+            s2_averages.append(row_avgs[1])
+            s3_averages.append(row_avgs[2])
+            
+        # Check s4 and s5
+        s4_ok = float(row[s4])<=1.0 and float(row[s4])>=0.0
+        s5_ok = float(row[s5])<=1.0 and float(row[s5])>=0.0
+        s6_ok = float(row[s6])<=1.0 and float(row[s6])>=0.0
+        if s4_ok and s5_ok and s6_ok:
+            s4_values.append(float(row[s4]))
+            s5_values.append(float(row[s5]))
+            s6_values.append(float(row[s6]))
+            s45_values.append((w1*float(row[s4]) + w2*float(row[s5])))
+        else:
+            invalid_s4_s5_s6_rows.append(idx+2)
+            print(type(float(row[s4])))
+            
+
+    if invalid_rows:
+        print(f"Invalid rows in 's1', 's2', or 's3': {invalid_rows}")
+    else:
+        print("All rows in 's1', 's2', and 's3' are valid.")
+
+    if invalid_s4_s5_s6_rows:
+        print(f"Invalid rows in 's4' or 's5': {invalid_s4_s5_s6_rows}")
+    else:
+        print("All rows in 's4' and 's5' are valid.")
+        
+    overall_s1_avg = sum(s1_averages) / len(s1_averages) if s1_averages else 0
+    overall_s2_avg = sum(s2_averages) / len(s2_averages) if s2_averages else 0
+    overall_s3_avg = sum(s3_averages) / len(s3_averages) if s3_averages else 0
+
+    print(f"Average for s1: {overall_s1_avg:.3f}")
+    print(f"Average for s2: {overall_s2_avg:.3f}")
+    print(f"Average for s3: {overall_s3_avg:.3f}")
+    
+    overall_s4_avg = sum(s4_values) / len(s4_values) if s4_values else 0
+    overall_s5_avg = sum(s5_values) / len(s5_values) if s5_values else 0
+    overall_s6_avg = sum(s6_values) / len(s6_values) if s6_values else 0
+    overall_s45_avg = sum(s45_values) / len(s45_values) if s45_values else 0
+    print(f"Average for s4: {overall_s4_avg:.3f}")
+    print(f"Average for s5: {overall_s5_avg:.3f}")
+    print(f"Average for s6: {overall_s6_avg:.3f}")
+    print(f"Average for s45: {overall_s45_avg:.3f}")
+    
+    return overall_s1_avg, overall_s2_avg, overall_s3_avg, overall_s4_avg, overall_s5_avg, overall_s6_avg,overall_s45_avg
+
+
+if __name__ == "__main__":
+    # model_names = [
+    #     "flux1-schnell",
+        
+    #     "sd35_large",
+    #     "sd35_medium",
+    #     "sd30_medium",
+        
+    #     "flux1-dev",
+    #     "playground-v25",
+    #     "hidream",
+        
+    #     "janus_pro_7B",
+    #     "show-o-512",
+    #     "bagel",
+    #     "emu3",
+    #     "GoT",
+        
+    #     "Gemini",
+    #     "GPT",
+    # ]
+    model_names = [
+            
+        # "hidream",
+        # "flux1-dev",
+        # "flux1-schnell",
+        # "playground-v25",
+        # "sd30_medium",
+        # "sd35_medium",
+        # "sd35_large",
+        
+        
+        # "emu3",
+        # "janus_pro_7B",
+        # "show-o-512",
+        # "GoT",
+        # "bagel",
+        
+        # "Gemini",
+        # "GPT",
+        
+        "qwen_image"
+    ]
+    s4_list = []
+    s5_list = []   
+    s6_list = []  
+    s45_list = [] 
+    for model_name in model_names:  
+        # csv_path = f'/group/xihuiliu/sky/reasoning/csv/idiom/{model_name}.csv'
+        # image_folder = f'/group/xihuiliu/sky/reasoning/images/idiom/{model_name}'
+        csv_path = f'/group/xihuiliu/sky/reasoning/csv/idiom/{model_name}_4o-pipeline.csv'
+        image_folder = f'/group/xihuiliu/sky/reasoning/models/GPT4o_pipeline/images_4o-pipeline/{model_name}'
+        
+        # csv_path = f'/group/xihuiliu/sky/reasoning/csv/text_image_new/{model_name}.csv'
+        # image_folder = f'/group/xihuiliu/sky/reasoning/images/text_image_new/{model_name}'
+        # csv_path = f'/group/xihuiliu/sky/reasoning/csv/text_image_new/{model_name}_4o-pipeline.csv'
+        # image_folder = f'/group/xihuiliu/sky/reasoning/models/GPT4o_pipeline/images_text_image_4o-pipeline/{model_name}'
+        
+        s1_column = 'score_alignment'
+        s2_column = 'score_quality'
+        s3_column = 'score_quality'
+        s4_column = 'score_a_avg'
+        s5_column = 'score_q_avg'
+        
+        s1,s2,s3,s4,s5=check(csv_path, image_folder, s1_column, s2_column, s3_column, s4_column, s5_column)  
+        s4_list.append(s4*100)
+        s5_list.append(s5*100)
+
+        
+        # csv_path = f'/group/xihuiliu/sky/reasoning/csv/entity/{model_name}_4o-pipeline.csv'
+        # image_folder = f'/group/xihuiliu/sky/reasoning/models/GPT4o_pipeline/images_entity_4o-pipeline/{model_name}'
+        # csv_path = f'/group/xihuiliu/sky/reasoning/csv/entity/{model_name}.csv'
+        # image_folder = f'/group/xihuiliu/sky/reasoning/images/common_sense/{model_name}'
+        
+        # s1_column = 'score_entity'
+        # s2_column = 'score_detail'
+        # s3_column = 'score_quality'
+        # s4_column = 'score_e_avg'
+        # s5_column = 'score_d_avg'
+        # s6_column = 'score_q_avg'
+        
+        # csv_path = f'/group/xihuiliu/sky/reasoning/csv/physics/{model_name}_4o-pipeline.csv'
+        # image_folder = f'/group/xihuiliu/sky/reasoning/models/GPT4o_pipeline/images_physics_4o-pipeline/{model_name}'
+        # csv_path = f'/group/xihuiliu/sky/reasoning/csv/physics/{model_name}.csv'
+        # image_folder = f'/group/xihuiliu/sky/reasoning/images/physics/{model_name}'
+        
+        # s1_column = 'score_scientific'
+        # s2_column = 'score_detail'
+        # s3_column = 'score_quality'
+        # s4_column = 'score_s_avg'
+        # s5_column = 'score_d_avg'
+        # s6_column = 'score_q_avg'
+        
+        # s1,s2,s3,s4,s5,s6,s45=check2(csv_path, image_folder, s1_column, s2_column, s3_column, s4_column, s5_column, s6_column)  
+        # s4_list.append(s4*100)
+        # s5_list.append(s5*100)
+        # s6_list.append(s6*100)
+        # s45_list.append(s45*100)
+
+for i, model_name in enumerate(model_names):
+            
+    print(f"{model_name},", round(s4_list[i],1), ",", round(s5_list[i],1))
+    # print(f"{model_name},", round(s4_list[i],1), ",", round(s5_list[i],1), ",", round(s6_list[i],1))
+    # print(f"{model_name},", round(s45_list[i],1), ",", round(s6_list[i],1))
+
+# idiom: GPT 198
+# hidream, 48.5 , 87.2
+# flux1-dev, 39.1 , 83.4
+# flux1-schnell, 40.9 , 83.1
+# playground-v25, 43.9 , 87.8
+# sd30_medium, 35.9 , 81.4
+# sd35_medium, 34.4 , 80.6
+# sd35_large, 35.6 , 85.3
+# emu3, 33.1 , 82.9
+# janus_pro_7B, 25.5 , 78.0
+# show-o-512, 33.1 , 82.5
+# GoT, 29.7 , 76.4
+# bagel, 44.6 , 84.3
+# Gemini, 52.4 , 87.8
+# 从133开始错误： GPT, 75.6 , 95.4, 更新：GPT, 75.7 , 94.5
+# qwen_image, 51.7 , 89.7
+
+#idiom pipeline
+# hidream, 64.4 , 91.9
+# flux1-dev, 66.2 , 90.5
+# flux1-schnell, 68.2 , 87.4
+# playground-v25, 55.8 , 88.7
+# sd30_medium, 65.7 , 87.6
+# sd35_medium, 66.8 , 88.5
+# sd35_large, 67.7 , 90.4
+# emu3, 56.0 , 84.2
+# janus_pro_7B, 63.1 , 82.9
+# show-o-512, 64.2 , 89.5
+# GoT, 51.8 , 81.4
+# bagel, 67.7 , 87.8
+# Gemini, 67.1 , 91.5
+# GPT, 77.3 , 93.8
+
+# entity pipeline
+# hidream, 76.4 , 77.9 , 96.8
+# 有错误：flux1-dev, 70.6 , 77.2 , 95.9  更新后：flux1-dev, 69.9 , 76.9 , 96.1
+# flux1-schnell, 70.1 , 78.4 , 94.9
+# playground-v25, 71.5 , 68.9 , 94.7
+# sd30_medium, 71.8 , 76.9 , 96.1
+# sd35_medium, 70.1 , 77.2 , 95.9
+# sd35_large, 76.8 , 79.8 , 95.6
+# emu3, 60.8 , 67.8 , 90.6
+# janus_pro_7B, 67.3 , 74.3 , 93.0
+# show-o-512, 64.6 , 70.9 , 94.0
+# GoT, 48.7 , 57.9 , 89.2
+# bagel, 66.9 , 76.8 , 94.7
+# Gemini, 77.9 , 77.8 , 96.1
+# GPT, 82.6 , 85.4 , 98.0
+#整体
+# hidream, 76.4 , 77.9 , 96.8
+# flux1-dev, 69.9 , 76.9 , 96.1
+# flux1-schnell, 70.1 , 78.4 , 94.9
+# playground-v25, 71.5 , 68.9 , 94.7
+# sd30_medium, 71.8 , 76.9 , 96.1
+# sd35_medium, 70.1 , 77.2 , 95.9
+# sd35_large, 76.8 , 79.8 , 95.6
+# emu3, 60.8 , 67.8 , 90.6
+# janus_pro_7B, 67.3 , 74.3 , 93.0
+# show-o-512, 64.6 , 70.9 , 94.0
+# GoT, 48.7 , 57.9 , 89.2
+# bagel, 66.9 , 76.8 , 94.7
+# Gemini, 77.9 , 77.8 , 96.1
+# GPT, 82.6 , 85.4 , 98.0
+#!!!整体 0.7 0.3
+# hidream, 76.9 , 96.8
+# flux1-dev, 72.0 , 96.1
+# flux1-schnell, 72.6 , 94.9
+# playground-v25, 70.7 , 94.7
+# sd30_medium, 73.3 , 96.1
+# sd35_medium, 72.2 , 95.9
+# sd35_large, 77.7 , 95.6
+# emu3, 62.9 , 90.6
+# janus_pro_7B, 69.4 , 93.0
+# show-o-512, 66.5 , 94.0
+# GoT, 51.5 , 89.2
+# bagel, 69.9 , 94.7
+# Gemini, 77.9 , 96.1
+# GPT, 83.4 , 98.0
+
+# entity
+# hidream, 47.1 , 70.4 , 94.1
+# 有错误：flux1-dev, 40.7 , 59.8 , 90.1 更新后：flux1-dev, 39.2 , 58.7 , 90.6
+# flux1-schnell, 37.5 , 62.0 , 91.5
+# playground-v25, 43.0 , 60.9 , 92.4
+# sd30_medium, 36.5 , 56.3 , 90.1
+# sd35_medium, 37.9 , 60.8 , 92.1
+# sd35_large, 40.7 , 60.4 , 92.6
+# emu3, 26.1 , 51.8 , 85.2
+# janus_pro_7B, 31.4 , 55.1 , 87.6
+# 有错误,一个是null show-o-512, 28.2 , 50.5 , 87.4
+# GoT, 25.8 , 43.1 , 86.2
+# bagel, 47.9 , 63.0 , 91.6
+# 从49开始错误： Gemini, 14.2 , 21.9 , 70.4  更新后：Gemini, 66.0 , 69.3 , 94.3
+# 从54开始错误： GPT, 19.2 , 24.3 , 70.1     更新后：GPT, 76.2 , 80.3 , 96.6
+#整体
+# hidream, 47.1 , 70.4 , 94.1
+# flux1-dev, 39.2 , 58.7 , 90.6
+# flux1-schnell, 37.5 , 62.0 , 91.5
+# playground-v25, 43.0 , 60.9 , 92.4
+# sd30_medium, 36.5 , 56.3 , 90.1
+# sd35_medium, 37.9 , 60.8 , 92.1
+# sd35_large, 40.7 , 60.4 , 92.6
+# emu3, 26.1 , 51.8 , 85.2
+# janus_pro_7B, 31.4 , 55.1 , 87.6
+# show-o-512, 28.2 , 50.5 , 87.4
+# GoT, 25.8 , 43.1 , 86.2
+# bagel, 47.9 , 63.0 , 91.6
+# Gemini, 66.0 , 69.3 , 94.3
+# GPT, 76.2 , 80.3 , 96.6
+#整体 0.7 0.3
+# hidream, 54.1 , 94.1
+# flux1-dev, 45.1 , 90.6
+# flux1-schnell, 44.8 , 91.5
+# playground-v25, 48.4 , 92.4
+# sd30_medium, 42.4 , 90.1
+# sd35_medium, 44.8 , 92.1
+# sd35_large, 46.6 , 92.6
+# emu3, 33.8 , 85.2
+# janus_pro_7B, 38.5 , 87.6
+# show-o-512, 34.9 , 87.4
+# GoT, 31.0 , 86.2
+# bagel, 52.4 , 91.6
+# Gemini, 67.0 , 94.3
+# GPT, 77.5 , 96.6
+# qwen_image, 60.6 , 96.2
+
+
+# text-image
+# hidream, 72.3 , 85.5
+# flux1-dev, 56.9 , 76.5
+# flux1-schnell, 65.1 , 74.5
+# 有错误,一个是null：playground-v25, 38.5 , 72.1   
+# sd30_medium, 60.9 , 71.3
+# sd35_medium, 58.0 , 70.1
+# sd35_large, 62.2 , 75.4
+# 有错误,一个是null：emu3, 33.7 , 68.7            
+# janus_pro_7B, 37.2 , 70.9
+# show-o-512, 35.3 , 80.3
+# 从3开始错误：GoT, 7.2 , 56.9, 更新后： GoT, 30.6 , 70.7
+# bagel, 44.0 , 73.7
+# 从53开始错误：Gemini, 28.0 , 70.1 更新后：Gemini, 73.0 , 83.3
+# GPT, 86.9 , 97.6
+# qwen_image, 72.0 , 81.3
+
+# text-image pipeline
+# hidream, 77.5 , 87.0
+# flux1-dev, 69.8 , 80.5
+# flux1-schnell, 71.6 , 78.7
+# playground-v25, 40.5 , 76.5
+# sd30_medium, 70.9 , 83.1
+# sd35_medium, 69.2 , 79.9
+# sd35_large, 72.4 , 84.4
+# emu3, 41.5 , 74.7
+# janus_pro_7B, 54.9 , 80.5
+# show-o-512, 42.9 , 83.5
+# 从177开始错误: GoT, 36.4 , 73.1   更新后：GoT, 36.4 , 73.1
+# 有错误：bagel, 61.0 , 79.1   更新后： bagel, 61.5 , 79.7
+# 从20开始错误: Gemini, 20.7 , 74.9 更新后： Gemini, 78.4 , 89.4
+# 从85开始错误: GPT, 41.7 , 86.1 更新后： GPT, 83.0 , 97.5
+
+#scientific
+# hidream, 45.0 , 72.5
+# 有错误: flux1-dev, 38.8 , 64.9  更新：flux1-dev, 38.8 , 65.1 , 80.9
+# flux1-schnell, 41.2 , 73.0
+# 有错误: playground-v25, 42.8 , 67.2, 更新：playground-v25, 43.6 , 67.5 , 83.3
+# sd30_medium, 42.3 , 71.0
+# sd35_medium, 42.8 , 66.4
+# sd35_large, 44.6 , 72.3
+# emu3, 33.8 , 54.7
+# janus_pro_7B, 37.1 , 63.3
+# show-o-512, 32.9 , 62.0
+# GoT, 30.8 , 50.9
+# bagel, 52.1 , 70.9
+# 从25开始错误: Gemini, 13.4 , 22.9 更新后：
+# 从104开始错误: GPT, 42.2 , 57.5   更新后：
+#整体
+# hidream, 45.0 , 72.5 , 84.5
+# flux1-dev, 38.8 , 65.1 , 80.9
+# flux1-schnell, 41.2 , 73.0 , 83.0
+# playground-v25, 43.6 , 67.5 , 83.3
+# sd30_medium, 42.3 , 71.0 , 81.7
+# sd35_medium, 42.8 , 66.4 , 83.0
+# sd35_large, 44.6 , 72.3 , 84.5
+# emu3, 33.8 , 54.7 , 77.0
+# janus_pro_7B, 37.1 , 63.3 , 77.8
+# show-o-512, 32.9 , 62.0 , 76.6
+# GoT, 30.8 , 50.9 , 76.3
+# bagel, 52.1 , 70.9 , 88.3
+# Gemini, 60.7 , 80.7 , 89.3
+# GPT, 68.7 , 88.5 , 94.3
+# ！！！整体 0.7 0.3
+# hidream, 53.2 , 84.5
+# flux1-dev, 46.7 , 80.9
+# flux1-schnell, 50.7 , 83.0
+# playground-v25, 50.8 , 83.3
+# sd30_medium, 50.9 , 81.7
+# sd35_medium, 49.9 , 83.0
+# sd35_large, 52.9 , 84.5
+# emu3, 40.1 , 77.0
+# janus_pro_7B, 44.9 , 77.8
+# show-o-512, 41.6 , 76.6
+# GoT, 36.8 , 76.3
+# bagel, 57.7 , 88.3
+# Gemini, 66.7 , 89.3
+# GPT, 74.7 , 94.3
+# qwen_image, 61.0 , 87.5
+
+
+# physics pipeline
+# hidream, 63.3 , 76.6
+# 有错误：flux1-dev, 64.4 , 79.1 更新后，有一个测不出来【149】：64.8 , 79.2 , 92.3
+# 有错误,40测不出来：flux1-schnell, 62.5 , 75.5, 更新：flux1-schnell, 62.2 , 75.4
+# playground-v25, 52.5 , 64.8
+# sd30_medium, 60.6 , 77.1
+# sd35_medium, 63.2 , 74.4
+# sd35_large, 65.2 , 76.4
+# janus_pro_7B, 56.7 , 71.4
+# show-o-512, 55.9 , 69.2
+# GoT, 38.7 , 54.9
+# bagel, 63.6 , 76.7
+# GPT, 78.8 , 85.4
+# 整体
+# hidream, 63.3 , 76.6 , 89.9
+# flux1-dev, 64.8 , 79.2 , 92.3
+# flux1-schnell, 62.2 , 75.4 , 90.1
+# playground-v25, 52.5 , 64.8 , 87.0
+# sd30_medium, 60.6 , 77.1 , 91.7
+# sd35_medium, 63.2 , 74.4 , 89.9
+# sd35_large, 65.2 , 76.4 , 92.3
+# emu3, 44.7 , 57.4 , 84.1
+# janus_pro_7B, 56.7 , 71.4 , 87.5
+# show-o-512, 55.9 , 69.2 , 90.3
+# GoT, 38.7 , 54.9 , 81.8
+# bagel, 63.6 , 76.7 , 90.3
+# Gemini, 69.2 , 78.9 , 90.6
+# GPT, 78.8 , 85.4 , 95.4
+# !!! 整体 0.7, 0.3
+# hidream, 67.3 , 89.9
+# flux1-dev, 69.1 , 92.3
+# flux1-schnell, 66.1 , 90.1
+# playground-v25, 56.1 , 87.0
+# sd30_medium, 65.5 , 91.7
+# sd35_medium, 66.6 , 89.9
+# sd35_large, 68.6 , 92.3
+# emu3, 48.5 , 84.1
+# janus_pro_7B, 61.1 , 87.5
+# show-o-512, 59.9 , 90.3
+# GoT, 43.6 , 81.8
+# bagel, 67.5 , 90.3
+# Gemini, 72.1 , 90.6
+# GPT, 80.8 , 95.4

csv/idiom/csv_tidyup.py

@@ -0,0 +1,27 @@
+import csv
+
+input_file = '/group/xihuiliu/sky/reasoning/csv/physics/nano_banana.csv'
+output_file = '/group/xihuiliu/sky/reasoning/csv/physics/nano_banana_clean.csv'
+
+# Read CSV and keep only the first occurrence of each image id
+unique_rows = {}
+with open(input_file, newline='', encoding='utf-8') as csvfile:
+    reader = csv.reader(csvfile)
+    header = next(reader)
+    for row in reader:
+        image_id = row[0]
+        if image_id not in unique_rows:
+            unique_rows[image_id] = row
+
+# Sort image ids from 0001.png to 0200.png
+sorted_image_ids = sorted(
+    unique_rows.keys(),
+    key=lambda x: int(x.split('.')[0])
+)
+sorted_rows = [unique_rows[img_id] for img_id in sorted_image_ids if 1 <= int(img_id.split('.')[0]) <= 200]
+
+# Write cleaned and sorted data to new CSV
+with open(output_file, 'w', newline='', encoding='utf-8') as csvfile:
+    writer = csv.writer(csvfile)
+    writer.writerow(header)
+    writer.writerows(sorted_rows)