Trainer - deprecate tokenizer for processing_class (#32385)

* Trainer - deprecate tokenizer for processing_class

* Extend chage across Seq2Seq trainer and docs

* Add tests

* Update to FutureWarning and add deprecation version

tests/trainer/test_trainer_seq2seq.py

@@ -129,7 +129,7 @@ class Seq2seqTrainerTester(TestCasePlus):
             compute_metrics=_compute_metrics,
             train_dataset=train_dataset,
             eval_dataset=val_dataset,
-            tokenizer=tokenizer,
+            processing_class=tokenizer,
         )
 
         # start training
@@ -158,7 +158,7 @@ class Seq2seqTrainerTester(TestCasePlus):
         trainer = Seq2SeqTrainer(
             model=model,
             args=training_args,
-            tokenizer=tokenizer,
+            processing_class=tokenizer,
             data_collator=data_collator,
             compute_metrics=lambda x: {"samples": x[0].shape[0]},
         )
@@ -199,7 +199,7 @@ class Seq2seqTrainerTester(TestCasePlus):
             _ = Seq2SeqTrainer(
                 model=model,
                 args=training_args,
-                tokenizer=tokenizer,
+                processing_class=tokenizer,
                 data_collator=data_collator,
                 compute_metrics=lambda x: {"samples": x[0].shape[0]},
             )