Fix block_size picking in megatron_lm_gpt_pretraining.py (#2342)

Only cap `block_size` to 1024 if `tokenizer.model_max_length` is actually greater than 1024.

examples/by_feature/megatron_lm_gpt_pretraining.py

@@ -405,7 +405,7 @@ def tokenize_function(examples):
                 f"The tokenizer picked seems to have a very large `model_max_length` ({tokenizer.model_max_length}). "
                 "Picking 1024 instead. You can change that default value by passing --block_size xxx."
             )
-        block_size = 1024
+            block_size = 1024
     else:
         if args.block_size > tokenizer.model_max_length:
             logger.warning(