diff --git a/lmms_eval/models/internvl2.py b/lmms_eval/models/internvl2.py
index 5f4365d0..509df583 100644
--- a/lmms_eval/models/internvl2.py
+++ b/lmms_eval/models/internvl2.py
@@ -145,6 +145,7 @@ def __init__(
 
         accelerator_kwargs = InitProcessGroupKwargs(timeout=timedelta(weeks=52))
         accelerator = Accelerator(kwargs_handlers=[accelerator_kwargs])
+        self.accelerator = accelerator
         if accelerator.num_processes > 1:
             self._device = torch.device(f"cuda:{accelerator.local_process_index}")
             self.device_map = f"cuda:{accelerator.local_process_index}"
diff --git a/lmms_eval/models/llava.py b/lmms_eval/models/llava.py
index 7d6420ba..8537f999 100755
--- a/lmms_eval/models/llava.py
+++ b/lmms_eval/models/llava.py
@@ -69,6 +69,7 @@ def __init__(
 
         accelerator_kwargs = InitProcessGroupKwargs(timeout=timedelta(weeks=52))
         accelerator = Accelerator(kwargs_handlers=[accelerator_kwargs])
+        self.accelerator = accelerator
         if accelerator.num_processes > 1:
             self._device = torch.device(f"cuda:{accelerator.local_process_index}")
             self.device_map = f"cuda:{accelerator.local_process_index}"
diff --git a/lmms_eval/models/longva.py b/lmms_eval/models/longva.py
index c5bf6861..040c551d 100644
--- a/lmms_eval/models/longva.py
+++ b/lmms_eval/models/longva.py
@@ -76,6 +76,7 @@ def __init__(
 
         accelerator_kwargs = InitProcessGroupKwargs(timeout=timedelta(weeks=52))
         accelerator = Accelerator(kwargs_handlers=[accelerator_kwargs])
+        self.accelerator = accelerator
         if accelerator.num_processes > 1:
             self._device = torch.device(f"cuda:{accelerator.local_process_index}")
             self.device_map = f"cuda:{accelerator.local_process_index}"