fix: fire model

src/model/model_llava.py

@@ -102,9 +102,9 @@ def inference_by_prompt_and_images_fire(prompt, images):
         for image in images:
             image_data.append(Image.open(BytesIO(base64.b64decode(image))))
         images = image_data
-    image_tensor = process_images(images, image_processor_llava, model_llava.config)
+    image_tensor = process_images(images, image_processor_llava_fire, model_llava_fire.config)
     image_tensor = image_tensor.to(dtype=torch.float16, device=device)
-    input_ids = tokenizer_image_token(prompt, tokenizer_llava, IMAGE_TOKEN_INDEX, return_tensors="pt").unsqueeze(0).to(device)
+    input_ids = tokenizer_image_token(prompt, tokenizer_llava_fire, IMAGE_TOKEN_INDEX, return_tensors="pt").unsqueeze(0).to(device)
     image_sizes = [image.size for image in images]
     logger.info("Shape: {};{}; Devices: {};{}",input_ids.shape, image_tensor.shape, input_ids.device, image_tensor.device)
     with torch.inference_mode():
@@ -117,7 +117,7 @@ def inference_by_prompt_and_images_fire(prompt, images):
             max_new_tokens=256,
             use_cache=True
         )
-    text_outputs = tokenizer_llava.batch_decode(cont, skip_special_tokens=True)
+    text_outputs = tokenizer_llava_fire.batch_decode(cont, skip_special_tokens=True)
     logger.info("response={}", text_outputs)
     return text_outputs