Fix --resume_from_checkpoint step in train_text_to_image.py (#1914)

fix resume step in train_text_to_image example

examples/text_to_image/train_text_to_image.py

@@ -685,9 +685,8 @@ def main():
         accelerator.load_state(os.path.join(args.output_dir, path))
         global_step = int(path.split("-")[1])
 
-        resume_global_step = global_step * args.gradient_accumulation_steps
-        first_epoch = resume_global_step // num_update_steps_per_epoch
-        resume_step = resume_global_step % num_update_steps_per_epoch
+        first_epoch = global_step // num_update_steps_per_epoch
+        resume_step = global_step % num_update_steps_per_epoch
 
     # Only show the progress bar once on each machine.
     progress_bar = tqdm(range(global_step, args.max_train_steps), disable=not accelerator.is_local_main_process)