Update app.py

app.py

@@ -28,7 +28,8 @@ class Settings:
     AGGREGATOR_URL = os.getenv("AGGREGATOR_URL", "http://192.168.1.104:8002")
     
     # Model settings
-    MODEL_REPO = "https://huggingface.co/microsoft/florence-2-large"
+    MODEL_REPO = "https://huggingface.co/facebook/opt-125m"
+    
     # Server settings
     TENSOR_SERVER_TIMEOUT = 30  # seconds
     MAX_ERROR_THRESHOLD = 5     # maximum number of errors
@@ -159,7 +160,24 @@ async def split_model_weights():
                 raise Exception("No model weight files found")
 
         # Get file size and calculate chunks
-        file_size = os.path.getsize(model_file)
+        try:
+            with open(model_file, 'rb') as f:
+                # Get actual file size by seeking to end
+                f.seek(0, 2)  # Seek to end
+                file_size = f.tell()  # Get position (total size)
+                f.seek(0)  # Reset to beginning
+                
+                # Read first few bytes to verify file isn't corrupted
+                header = f.read(8)
+                if len(header) == 0:
+                    raise ValueError(f"File is empty: {model_file}")
+        except Exception as e:
+            raise Exception(f"Failed to read model file {model_file}: {str(e)}")
+            
+        # Verify file size is reasonable
+        if file_size < 1024:  # Less than 1KB
+            raise ValueError(f"Model file suspiciously small ({file_size} bytes). Possible corruption or incomplete download.")
+            
         num_servers = len(state.tensor_servers) or len(Settings.TENSOR_SERVER_URLS)
         num_chunks = num_servers  # One chunk per server initially
         
@@ -168,13 +186,13 @@ async def split_model_weights():
         # Format sizes for display
         def format_size(size_bytes):
             if size_bytes >= 1024*1024*1024:  # GB
-                return f"{size_bytes / (1024*1024*1024):.2f} GB"
+                return f"{size_bytes / (1024*1024*1024):.2f} GB ({size_bytes:,} bytes)"
             elif size_bytes >= 1024*1024:  # MB
-                return f"{size_bytes / (1024*1024):.2f} MB"
+                return f"{size_bytes / (1024*1024):.2f} MB ({size_bytes:,} bytes)"
             elif size_bytes >= 1024:  # KB
-                return f"{size_bytes / 1024:.2f} KB"
+                return f"{size_bytes / 1024:.2f} KB ({size_bytes:,} bytes)"
             else:
-                return f"{size_bytes} bytes"
+                return f"{size_bytes:,} bytes"
         
         print(f"[INFO] Model file size: {format_size(file_size)}")
         print(f"[INFO] Creating {num_chunks} chunks of approximately {format_size(chunk_size)} each")
@@ -197,13 +215,21 @@ async def split_model_weights():
                     break
                 
                 # Read and write chunk
-                f.seek(start_pos)
-                chunk_data = f.read(current_chunk_size)
-                
-                with open(chunk_path, 'wb') as chunk_file:
-                    chunk_file.write(chunk_data)
-                
-                chunk_sizes.append(current_chunk_size)
+                try:
+                    f.seek(start_pos)
+                    chunk_data = f.read(current_chunk_size)
+                    actual_chunk_size = len(chunk_data)
+                    
+                    if actual_chunk_size != current_chunk_size:
+                        print(f"[WARN] Chunk {chunk_id} size mismatch. Expected: {current_chunk_size}, Got: {actual_chunk_size}")
+                    
+                    with open(chunk_path, 'wb') as chunk_file:
+                        chunk_file.write(chunk_data)
+                    
+                    chunk_sizes.append(actual_chunk_size)
+                    print(f"[DEBUG] Chunk {chunk_id} data: First few bytes: {chunk_data[:20].hex()}")
+                except Exception as e:
+                    raise Exception(f"Failed to process chunk {chunk_id} at offset {start_pos}: {str(e)}")
                 
                 # Create chunk metadata
                 state.model_chunks[chunk_id] = ModelChunk(
@@ -1065,7 +1091,7 @@ if __name__ == "__main__":
     print(f"[INFO] API Documentation available at http://localhost:{port}/docs")
     
     uvicorn.run(
-        "controller_server_new:app",
+        "app:app",
         host="0.0.0.0",
         port=port,
         reload=False