remove offload_state_dict from kwargs

Cyrilvallez · Cyrilvallez · commit a92b1e8a45e1 · 2025-10-14T16:46:22.000+02:00
diff --git a/src/transformers/modeling_utils.py b/src/transformers/modeling_utils.py
@@ -4667,6 +4667,7 @@ def from_pretrained(
         _ = kwargs.pop("mirror", None)
         _ = kwargs.pop("_fast_init", True)
         _ = kwargs.pop("low_cpu_mem_usage", None)
+        _ = kwargs.pop("offload_state_dict", None)
 
         # For BC on torch_dtype argument
         if torch_dtype is not None:
