diff --git a/src/accelerate/accelerator.py b/src/accelerate/accelerator.py
index b17ea387d54..552c8392c08 100644
--- a/src/accelerate/accelerator.py
+++ b/src/accelerate/accelerator.py
@@ -747,6 +747,12 @@ def _prepare_deepspeed(self, *args):
             batch_size_per_device = deepspeed_plugin.deepspeed_config["train_micro_batch_size_per_gpu"]
             result = [obj for obj in args]
 
+        if self.gradient_accumulation_steps != deepspeed_plugin.deepspeed_config["gradient_accumulation_steps"]:
+            logger.info(
+                f"Updating DeepSpeed's gradient accumulation steps to {self.gradient_accumulation_steps} from "
+                f"{deepspeed_plugin.deepspeed_config['gradient_accumulation_steps']}."
+            )
+            deepspeed_plugin.deepspeed_config["gradient_accumulation_steps"] = self.gradient_accumulation_steps
         config_kwargs = {
             "train_micro_batch_size_per_gpu": batch_size_per_device,
             "train_batch_size": batch_size_per_device