Add assert for pp and untie

Signed-off-by: MaximumEntropy <sandeep.subramanian.1@umontreal.ca>

megatron/arguments.py

@@ -349,6 +349,7 @@ def validate_args(args, defaults={}):
                 "Using async gradient all reduce requires setting the environment "
                 "variable CUDA_DEVICE_MAX_CONNECTIONS to 1")
 
+
     # Load retro args.
     if args.retro_workdir:
         retro_args_path = get_retro_args_path(args.retro_workdir)
@@ -367,6 +368,7 @@ def validate_args(args, defaults={}):
     if retro_args and args != retro_args:
         _print_args("retro arguments", types.SimpleNamespace(**{k:v for k,v in vars(retro_args).items() if k.startswith("retro")}, rank=args.rank))
 
+
     return args