update

eagle_vl/serve/inference.py

@@ -44,6 +44,8 @@ def load_model_from_nv(model_path: str = "nvidia/Eagle-2.5-8B"):
 def load_model_from_eagle(model_path: str = "NVEagle/Eagle2.5-VL-8B-Preview"):
 
     token = os.environ.get("HF_TOKEN")
+    logger.info(f"token = {token[:4]}***{token[-2:]}")
+
     # hotfix the model to use flash attention 2
     config = AutoConfig.from_pretrained(model_path, trust_remote_code=True, token=token)
     config._attn_implementation = "flash_attention_2"
@@ -51,7 +53,6 @@ def load_model_from_eagle(model_path: str = "NVEagle/Eagle2.5-VL-8B-Preview"):
     config.text_config._attn_implementation = "flash_attention_2"
     print("Successfully set the attn_implementation to flash_attention_2")
 
-    logger.info(f"token = {token[:4]}***{token[-2:]}")
     model = AutoModel.from_pretrained(
         model_path,
         trust_remote_code=True,