Add torch `RMSProp` optimizer (#26425)

add rmsprop

src/transformers/trainer.py

@@ -1139,6 +1139,8 @@ class Trainer:
             optimizer_cls = torch.optim.SGD
         elif args.optim == OptimizerNames.ADAGRAD:
             optimizer_cls = torch.optim.Adagrad
+        elif args.optim == OptimizerNames.RMSPROP:
+            optimizer_cls = torch.optim.RMSprop
         else:
             raise ValueError(f"Trainer cannot instantiate unsupported optimizer: {args.optim}")
         return optimizer_cls, optimizer_kwargs

src/transformers/training_args.py

@@ -153,6 +153,7 @@ class OptimizerNames(ExplicitEnum):
     PAGED_ADAMW_8BIT = "paged_adamw_8bit"
     PAGED_LION = "paged_lion_32bit"
     PAGED_LION_8BIT = "paged_lion_8bit"
+    RMSPROP = "rmsprop"
 
 
 # TODO: `TrainingArguments` users rely on it being fully mutable. In the future see if we can narrow this to a few keys: https://github.com/huggingface/transformers/pull/25903