Fix error with global step in run_lm_finetuning.py

examples/run_lm_finetuning.py

@@ -264,8 +264,10 @@ def train(args, train_dataset, model, tokenizer):
     steps_trained_in_current_epoch = 0
     # Check if continuing training from a checkpoint
     if os.path.exists(args.model_name_or_path):
+        try:
             # set global_step to gobal_step of last saved checkpoint from model path
-        global_step = int(args.model_name_or_path.split("-")[-1].split("/")[0])
+            checkpoint_suffix = args.model_name_or_path.split("-")[-1].split("/")[0]
+            global_step = int(checkpoint_suffix)
             epochs_trained = global_step // (len(train_dataloader) // args.gradient_accumulation_steps)
             steps_trained_in_current_epoch = global_step % (len(train_dataloader) // args.gradient_accumulation_steps)
 
@@ -273,6 +275,8 @@ def train(args, train_dataset, model, tokenizer):
             logger.info("  Continuing training from epoch %d", epochs_trained)
             logger.info("  Continuing training from global step %d", global_step)
             logger.info("  Will skip the first %d steps in the first epoch", steps_trained_in_current_epoch)
+        except ValueError:
+            logger.info("  Starting fine-tuning.")
 
     tr_loss, logging_loss = 0.0, 0.0