[fxi] get_inference_model_by_args: use args_to_parse argument

app.py

@@ -239,12 +239,12 @@ def get_inference_model_by_args(args_to_parse):
             return output_image, output_str
 
         # Model Inference
-        conv = conversation_lib.conv_templates[args.conv_type].copy()
+        conv = conversation_lib.conv_templates[args_to_parse.conv_type].copy()
         conv.messages = []
 
         prompt = input_str
         prompt = DEFAULT_IMAGE_TOKEN + "\n" + prompt
-        if args.use_mm_start_end:
+        if args_to_parse.use_mm_start_end:
             replace_token = (
                 DEFAULT_IM_START_TOKEN + DEFAULT_IMAGE_TOKEN + DEFAULT_IM_END_TOKEN
             )
@@ -265,9 +265,9 @@ def get_inference_model_by_args(args_to_parse):
             .unsqueeze(0)
             .cuda()
         )
-        if args.precision == "bf16":
+        if args_to_parse.precision == "bf16":
             image_clip = image_clip.bfloat16()
-        elif args.precision == "fp16":
+        elif args_to_parse.precision == "fp16":
             image_clip = image_clip.half()
         else:
             image_clip = image_clip.float()
@@ -280,9 +280,9 @@ def get_inference_model_by_args(args_to_parse):
             .unsqueeze(0)
             .cuda()
         )
-        if args.precision == "bf16":
+        if args_to_parse.precision == "bf16":
             image = image.bfloat16()
-        elif args.precision == "fp16":
+        elif args_to_parse.precision == "fp16":
             image = image.half()
         else:
             image = image.float()