update

Author: sabulin

examples/OmniGen2-RL/README.md

@@ -72,7 +72,11 @@ Use EditScore to provide a high-quality reward signal to train models for signif
 
 ### 1. Data and Model Download
 Download RL training data from [EditScore-RL-Data](https://huggingface.co/datasets/EditScore/EditScore-RL-Data), then put the `rl.jsonl` into `data/` and change its path in `data_configs/train/train.yml`
-
+To convert relative image paths to your absolute paths:
+```bash
+cd examples/OmniGen2-RL/data
+python process_jsonl.py --input path/to/input.jsonl --output path/to/output.jsonl --base-path /your/absolute/base/path
+```
 Download the base model OmniGen2 form [OmniGen2](https://huggingface.co/OmniGen2/OmniGen2),then change the model file format to pytorch_model.bin and modify `model.pretrained_model_path` in `options/omnigen2_edit_rl.yml`
 
 ### 2. Start Reward Server

examples/OmniGen2-RL/data/process_jsonl.py

@@ -0,0 +1,91 @@
+import json
+import os
+import argparse
+
+def convert_jsonl_paths(input_file, output_file, base_path, max_records=100000):
+    """
+    Convert JSONL file image paths from relative to absolute paths
+    
+    Args:
+        input_file: Path to input JSONL file
+        output_file: Path to output JSONL file
+        base_path: Base directory path for converting relative paths
+        max_records: Maximum number of records to process, default 10000
+    """
+    
+    with open(input_file, 'r', encoding='utf-8') as infile, \
+         open(output_file, 'w', encoding='utf-8') as outfile:
+        
+        count = 0
+        for line in infile:
+            if count >= max_records:
+                break
+                
+            # Parse JSON line
+            try:
+                data = json.loads(line.strip())
+                
+                # Check if input_images field exists
+                if "input_images" in data and isinstance(data["input_images"], list):
+                    # Convert paths
+                    new_paths = []
+                    for path in data["input_images"]:
+                        if isinstance(path, str) and path.startswith("images/"):
+                            # Convert relative path to absolute path
+                            new_path = path.replace("images/", f"{base_path}/images/")
+                            new_paths.append(new_path)
+                        else:
+                            # Keep original path unchanged
+                            new_paths.append(path)
+                    
+                    data["input_images"] = new_paths
+                
+                # Write converted data
+                outfile.write(json.dumps(data, ensure_ascii=False) + '\n')
+                count += 1
+                
+                # Print progress every 1000 records
+                if count % 1000 == 0:
+                    print(f"Processed {count} records")
+                    
+            except json.JSONDecodeError as e:
+                print(f"Skipping invalid JSON line: {e}")
+                continue
+    
+    print(f"Conversion completed! Total processed records: {count}")
+    print(f"Output file: {output_file}")
+
+def main():
+    parser = argparse.ArgumentParser(description="Convert JSONL file image paths from relative to absolute")
+    parser.add_argument("--input", "-i", required=True, help="Input JSONL file path")
+    parser.add_argument("--output", "-o", required=True, help="Output JSONL file path")
+    parser.add_argument("--base-path", "-b", required=True, help="Base directory path for converting relative paths")
+    parser.add_argument("--max-records", "-m", type=int, default=100000, help="Maximum number of records to process (default: 100000)")
+    
+    args = parser.parse_args()
+    
+    # Validate input file exists
+    if not os.path.exists(args.input):
+        print(f"Error: Input file '{args.input}' does not exist")
+        return
+    
+    # Create output directory if it doesn't exist
+    output_dir = os.path.dirname(args.output)
+    if output_dir and not os.path.exists(output_dir):
+        os.makedirs(output_dir)
+        print(f"Created output directory: {output_dir}")
+    
+    # Validate base path exists
+    if not os.path.exists(args.base_path):
+        print(f"Warning: Base path '{args.base_path}' does not exist")
+    
+    print(f"Input file: {args.input}")
+    print(f"Output file: {args.output}")
+    print(f"Base path: {args.base_path}")
+    print(f"Max records: {args.max_records}")
+    print("-" * 50)
+    
+    convert_jsonl_paths(args.input, args.output, args.base_path, args.max_records)
+
+if __name__ == "__main__":
+    main()