v10a

improve_gainlora/src/run_t5.py

@@ -844,6 +844,7 @@ def main():
         model.gradient_checkpointing_enable(
             gradient_checkpointing_kwargs={"use_reentrant": False}
         )
+        model.enable_input_require_grads()
 
     world_size = int(os.environ.get("WORLD_SIZE", 1))
     training_args.step_per_epoch = math.ceil(len(raw_datasets["train"]) / training_args.per_device_train_batch_size / world_size / training_args.gradient_accumulation_steps)