batched loading

selfies_converter.py

@@ -3,44 +3,70 @@ import os
 import selfies as sf
 from tqdm import tqdm
 
-def convert_file(input_file, verbose=False):
+def convert_file(input_file, verbose=False, batch_size=10000):
     base_name = os.path.splitext(input_file)[0]
     selfies_file = f"{base_name}.selfies"
     alphabet_file = f"{base_name}.alph"
     
     alphabet = set()
-    
+    total_lines = 0
+    processed_lines = 0
+
+    # Count total lines if verbose
+    if verbose:
+        with open(input_file, 'r') as fin:
+            total_lines = sum(1 for _ in fin)
+
     with open(input_file, 'r') as fin, open(selfies_file, 'w') as fout:
-        lines = fin.readlines()
         if verbose:
-            lines = tqdm(lines, desc=f"Converting {input_file}")
-        
-        for line in lines:
-            smiles = line.strip()
-            try:
-                selfies = sf.encoder(smiles, strict=False)
-                fout.write(f"{selfies}\n")
-                alphabet.update(sf.split_selfies(selfies))
-            except sf.EncoderError as e:
+            pbar = tqdm(total=total_lines, desc=f"Converting {input_file}")
+
+        batch = []
+        for line in fin:
+            batch.append(line.strip())
+            if len(batch) == batch_size:
+                process_batch(batch, fout, alphabet)
+                processed_lines += len(batch)
                 if verbose:
-                    print(f"Failed to encode: {smiles}, {e}")
-                continue
-    
+                    pbar.update(len(batch))
+                batch = []
+
+        # Process remaining lines
+        if batch:
+            process_batch(batch, fout, alphabet)
+            processed_lines += len(batch)
+            if verbose:
+                pbar.update(len(batch))
+
+        if verbose:
+            pbar.close()
+
     with open(alphabet_file, 'w') as f:
         for symbol in sorted(alphabet):
             f.write(f"{symbol}\n")
     
     if verbose:
-        print(f"Conversion complete. SELFIES saved to {selfies_file}")
+        print(f"Conversion complete. Processed {processed_lines} lines.")
+        print(f"SELFIES saved to {selfies_file}")
         print(f"Alphabet saved to {alphabet_file}")
 
+def process_batch(batch, fout, alphabet):
+    for smiles in batch:
+        try:
+            selfies = sf.encoder(smiles)
+            fout.write(f"{selfies}\n")
+            alphabet.update(sf.split_selfies(selfies))
+        except sf.EncoderError:
+            continue
+
 def main():
     parser = argparse.ArgumentParser(description="SELFIES Converter")
     parser.add_argument("input_file", help="Input SMILES file (.smi)")
     parser.add_argument("-v", "--verbose", action="store_true", help="Increase output verbosity")
+    parser.add_argument("-b", "--batch-size", type=int, default=10000, help="Batch size for processing")
     args = parser.parse_args()
 
-    convert_file(args.input_file, args.verbose)
+    convert_file(args.input_file, args.verbose, args.batch_size)
 
 if __name__ == "__main__":
-    main()
+    main()