Update app.py

app.py

@@ -11,11 +11,11 @@ import gradio as gr
 import spaces
 import torch
 
-from builder import load_pretrained_model
 from llava.constants import IMAGE_TOKEN_INDEX
 from llava.constants import LOGDIR
 from llava.conversation import (default_conversation, conv_templates)
 from llava.mm_utils import KeywordsStoppingCriteria, tokenizer_image_token
+from llava.model.builder import load_pretrained_model
 from llava.utils import (build_logger, violates_moderation, moderation_msg)
 from taxonomy import wrap_taxonomy, default_taxonomy
 
@@ -502,7 +502,7 @@ Set the environment variable `model` to change the model:
         'LukasHug/LlavaGuard-13B-hf',
         'LukasHug/LlavaGuard-34B-hf', ]
     bits = int(os.getenv("bits", 16))
-    model = os.getenv("model", models[0])
+    model = os.getenv("model", models[1])
     available_devices = os.getenv("CUDA_VISIBLE_DEVICES", "0")
     model_path, model_name = model, model.split("/")[0]
     if api_key:
@@ -516,7 +516,7 @@ Set the environment variable `model` to change the model:
         model_path = '/common-repos/LlavaGuard/models/LlavaGuard-v1.1-7b-full/smid_and_crawled_v2_with_augmented_policies/json-v16/llava'
 
     print(f"Loading model {model_path}")
-    tokenizer, model, image_processor, context_len = load_pretrained_model(model_path, None, model_name, token=api_key)
+    tokenizer, model, image_processor, context_len = load_pretrained_model(model_path, None, model_name)
     model.config.tokenizer_model_max_length = 2048 * 2
 
     exit_status = 0