diff --git a/garak/__main__.py b/garak/__main__.py
index 55e257a0..7603c4e4 100644
--- a/garak/__main__.py
+++ b/garak/__main__.py
@@ -2,7 +2,6 @@
 
 import sys
 
-sys.path.append("/home/sda/tianhaoli/garak")
 from garak import cli
 
 
diff --git a/garak/generators/huggingface.py b/garak/generators/huggingface.py
index e216285e..a3d8a3bf 100644
--- a/garak/generators/huggingface.py
+++ b/garak/generators/huggingface.py
@@ -84,7 +84,6 @@ def __init__(self, name, do_sample=True, generations=10, device=0):
             model=name,
             do_sample=do_sample,
             device=device,
-            proxy={'http': 'http://127.0.0.1:7890/', 'https': 'http://127.0.0.1:7890/'}
         )
         self.deprefix_prompt = name in models_to_deprefix
         if _config.loaded:
@@ -306,7 +305,6 @@ def _call_model(self, prompt: str, generations_this_call: int = 1) -> List[str]:
             headers=self.headers,
             json=payload,
             timeout=(20, 90),  # (connect, read)
-            proxies={'http': 'http://127.0.0.1:7890/', 'https': 'http://127.0.0.1:7890/'}
         )
 
         if req_response.status_code == 503:
@@ -536,18 +534,14 @@ class LLaVA(Generator):
     ]
     
     def __init__(self, name="", generations=10):
-        proxies = {
-            'http': 'http://127.0.0.1:7890',
-            'https': 'http://127.0.0.1:7890'
-        }
         if name not in self.supported_models:
             raise ValueError(
                 f"Invalid modal name {name}, current support: {self.supported_models}."
             )
-        self.processor = LlavaNextProcessor.from_pretrained(name, proxies=proxies)
+        self.processor = LlavaNextProcessor.from_pretrained(name)
         self.model = LlavaNextForConditionalGeneration.from_pretrained(name, 
                                                                        torch_dtype=torch.float16, 
-                                                                       low_cpu_mem_usage=True, proxies=proxies)
+                                                                       low_cpu_mem_usage=True)
         if torch.cuda.is_available():
             self.model.to("cuda:0")  
         else: