diff --git a/quantize_streaming.py b/quantize_streaming.py
index a0f1f4b..5aecf2d 100644
--- a/quantize_streaming.py
+++ b/quantize_streaming.py
@@ -30,11 +30,22 @@ def streaming_quantize(model_path: str, output_path: str):
     output_path = Path(output_path)
     output_path.mkdir(parents=True, exist_ok=True)
 
-    config = AutoConfig.from_pretrained(model_path, trust_remote_code=True)
-
     import glob
     shards = sorted(glob.glob(f"{model_path}/*.safetensors"))
+    
+    # Rename existing 0-indexed files to 1-indexed
+    for existing in output_path.glob("*.safetensors"):
+        parts = existing.name.split("-")
+        if len(parts) >= 3 and existing.name.startswith("model-00000-"):
+            num = int(parts[1])
+            new_name = f"model-{num+1:05d}-of-{len(shards):05d}.safetensors"
+            new_path = output_path / new_name
+            existing.rename(new_path)
+            print(f"Renamed: {existing.name} -> {new_name}")
+    
     print(f"Found {len(shards)} shards\n")
+    
+    config = AutoConfig.from_pretrained(model_path, trust_remote_code=True)
 
     # Process 4 shards per GPU (8 total) for max parallelism
     shards_per_gpu = 4