llava: fix typo in arg vs args

Signed-off-by: Aaron Teo <[email protected]>

Author: taronaeo

examples/llava/convert_image_encoder_to_gguf.py

@@ -365,7 +365,7 @@ def get_non_negative_vision_feature_layers(v_hparams):
 
         # byteswaps v.head.ffn_up.bias for big-endian systems
         # see: https://github.com/ggml-org/llama.cpp/issues/12863
-        if name == "v.head.ffn_up.bias" and arg.bigendian:
+        if name == "v.head.ffn_up.bias" and args.bigendian:
             data = data.byteswap(inplace=True)
 
         fout.add_tensor(name, data)