[refactor] code cleaning

model/LISA.py

@@ -3,13 +3,8 @@ from typing import List
 import torch
 import torch.nn as nn
 import torch.nn.functional as F
-from transformers import BitsAndBytesConfig, CLIPVisionModel
 
-from utils.utils import (DEFAULT_IM_END_TOKEN, DEFAULT_IM_START_TOKEN,
-                         DEFAULT_IMAGE_PATCH_TOKEN)
-
-from .llava.model.language_model.llava_llama import (LlavaLlamaForCausalLM,
-                                                     LlavaLlamaModel)
+from .llava.model.language_model.llava_llama import (LlavaLlamaForCausalLM, LlavaLlamaModel)
 from .segment_anything import build_sam_vit_h
 
 

utils/app_helpers.py

@@ -220,8 +220,7 @@ def get_inference_model_by_args(args_to_parse):
         conv = conversation_lib.conv_templates[args_to_parse.conv_type].copy()
         conv.messages = []
 
-        prompt = input_str
-        prompt = utils.DEFAULT_IMAGE_TOKEN + "\n" + prompt
+        prompt = utils.DEFAULT_IMAGE_TOKEN + "\n" + input_str
         if args_to_parse.use_mm_start_end:
             replace_token = (
                     utils.DEFAULT_IM_START_TOKEN + utils.DEFAULT_IMAGE_TOKEN + utils.DEFAULT_IM_END_TOKEN