[deepspeed] fix resume from checkpoint (#11352)

This PR fixes a bug that most likely somehow got exposed (not caused) by https://github.com/huggingface/transformers/pull/11318 - surprisingly the same test worked just fine before that other PR.

src/transformers/trainer.py

@@ -1017,7 +1017,7 @@ class Trainer:
                         "yield to errors or unwanted behaviors."
                     )
 
-            if self.deepspeed:
+            if args.deepspeed:
                 # will be resumed in deepspeed_init
                 pass
             else: