fix: weights_only=False for safetensors loading

quantize_streaming.py

@@ -33,7 +33,7 @@ def streaming_quantize(model_path, output_path):
         
         # Load shard to CPU
         print("   Loading shard to CPU...")
-        shard_state_dict = torch.load(shard_file, map_location="cpu", weights_only=True)
+        shard_state_dict = torch.load(shard_file, map_location="cpu", weights_only=False)
         
         # Quantize Linear layers in this shard using both GPUs
         print("   Quantizing Linear layers (both GPUs)...")