Fix load of rng state for resuming training from checkpoint (#37162)

Co-authored-by: Marc Sun <[email protected]>

Author: winglian

src/transformers/trainer.py

@@ -3122,7 +3122,7 @@ def _load_rng_state(self, checkpoint):
                 return
 
         with safe_globals():
-            checkpoint_rng_state = torch.load(rng_file, weights_only=True)
+            checkpoint_rng_state = torch.load(rng_file)
         random.setstate(checkpoint_rng_state["python"])
         np.random.set_state(checkpoint_rng_state["numpy"])
         torch.random.set_rng_state(checkpoint_rng_state["cpu"])