fix typo (#854)

intelligent-machine-learning · Nov 28, 2023 · 28ce84c · 28ce84c
1 parent c84fd53
commit 28ce84c
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/atorch/docs/auto_accelerate_api.md b/atorch/docs/auto_accelerate_api.md
@@ -311,7 +311,7 @@ ATorch-defined arguments:
 - atorch_wrap_cls: tuple of submodule names or submodule type for fsdp to wrap.
 - atorch_size_based_min_num_params： wrap submoudule based on parameter size. Should not used with atorch_wrap_cls.
 - atorch_ignored_cls: tuple of submodule names or submodule type for fsdp to ignore (not sharded).
-- fsdp_wrap_params_outmost: if True, wrap trainable parameters together in an outmost fsdp wrap. You may get better performance for finetuning with a small percentage of trainable parameters, such as LORA.
+- wrap_trainable_outmost: if True, wrap trainable parameters together in an outmost fsdp wrap. You may get better performance for finetuning with a small percentage of trainable parameters, such as LORA.
 
 Recommended configurations
 ```