ready for open, adjust the local path

Author: hxhcreate

README.md

@@ -50,6 +50,7 @@ Our code support several archs
 - qwen2vl: for Qwen2-VL
 - mllama: for Llama3.2-Vision
 
+If you are using openai apis, please remember to specify the model_name and your customized api_key and api_base in [load_openai.py](./models/load_openai.py)
 
 First, download the [dataset](https://huggingface.co/datasets/Foreshhh/vlsbench) in huggingface and specify the downloaded dir as the `ROOT_DIR`.
 

eval.py

@@ -72,7 +72,7 @@ def eval_main(data, data_root, model_inference, output_path):
 if __name__ == "__main__":
     parser = argparse.ArgumentParser()
     parser.add_argument("--arch", type=str, default="llava")   # if you specify the openai model, you need to specify the api name in load_openai.py
-    parser.add_argument("--data_root", type=str, default='/mnt/hwfile/trustai/huxuhao/vlsbench')
+    parser.add_argument("--data_root", type=str, default='~/vlsbench')
     parser.add_argument("--output_dir", type=str, default='./outputs')
     args = parser.parse_args()
 

eval_utils.py

@@ -59,8 +59,6 @@ class VLSBENCHOUTPUT(BaseModel):
 def get_openai_structered_eval_answer(
     data_messages, model_name, response_class=VLSBENCHOUTPUT
 ):
-    os.environ["http_proxy"] = "http://10.1.20.57:23128"
-    os.environ["https_proxy"] = "http://10.1.20.57:23128"
     api_base = f"https://api.openai.com/v1"
     api_key = os.environ.get("OPENAI_API_KEY")
     client = OpenAI(api_key=api_key, base_url=api_base)

models/load_llava.py

@@ -34,7 +34,7 @@ def load_image(image_file):
 
 # load_model
 print(f"[INFO] Load llava from origin!")
-model_path = "/mnt/hwfile/trustai/huxuhao/models/llava-v1.5-7b"
+model_path = "liuhaotian/llava-v1.5-7b"
 model_name = get_model_name_from_path(model_path)
 tokenizer, model, image_processor, context_len = load_pretrained_model(
     model_path, None, model_name

models/load_llava_hf.py

@@ -18,7 +18,7 @@
 
 # load_model
 print(f"[INFO] Load llavaforcontionalgeneration")
-model_path = "/mnt/hwfile/trustai/huxuhao/models/llava-1.5-7b-hf"
+model_path = "llava_hf/llava-1.5-7b-hf"
 model = LlavaForConditionalGeneration.from_pretrained(model_path, torch_dtype=torch.float16, low_cpu_mem_usage=True, device_map='cuda')
 processor = LlavaProcessor.from_pretrained(model_path)
 

models/load_mllama.py

@@ -11,7 +11,7 @@
 
 # load model
 print(f"[INFO] Load MllamaForConditionalGeneration")
-model_path = "/mnt/hwfile/trustai/huxuhao/models/Llama-3.2-11B-Vision-Instruct"
+model_path = "meta-llama/Llama-3.2-11B-Vision-Instruct"
 model = MllamaForConditionalGeneration.from_pretrained(model_path, torch_dtype=torch.bfloat16, device_map='auto').to("cuda")
 processor = AutoProcessor.from_pretrained(model_path)
 

models/load_openai.py

@@ -17,19 +17,13 @@
 api_base = ""
 
 if model_name == 'local':
-    if "http_proxy" in os.environ:
-        del os.environ["http_proxy"]
-    if 'HTTP_PROXY' in os.environ:
-        del os.environ["HTTP_PROXY"]
     client = OpenAI(api_key=api_key, base_url=api_base)
     model_name = client.models.list().data[0].id      
 else: 
     assert model_name in CLOSE_API_NAME
     if not api_key and not api_base:    # not set, then default
         api_key = os.environ.get("OPENAI_API_KEY")
         api_base = f"https://api.openai.com/v1"
-        os.environ['http_proxy'] = "http://10.1.20.57:23128"
-        os.environ['https_proxy'] = "http://10.1.20.57:23128"
     client = OpenAI(api_key=api_key, base_url=api_base)
 print(model_name)
 

models/load_qwen2vl.py

@@ -11,7 +11,7 @@
 
 # load model
 print(f"[INFO] Load qwenvl2forconditionalgeneration")
-model_path = "/mnt/hwfile/trustai/huxuhao/models/Qwen2-VL-7B-Instruct"
+model_path = "Qwen/Qwen2-VL-7B-Instruct"
 model = Qwen2VLForConditionalGeneration.from_pretrained(model_path, torch_dtype=torch.bfloat16).to("cuda")
 processor = AutoProcessor.from_pretrained(model_path)