refactor: change default block_size (#26229)

* refactor: change default block_size

* fix: return tf to origin

* fix: change files to origin

* rebase

* rebase

* rebase

* rebase

* rebase

* rebase

* rebase

* rebase

* refactor: add min block_size to files

* reformat: add min block_size for run_clm tf

examples/pytorch/language-modeling/run_clm.py

@@ -497,13 +497,12 @@ def main():
 
     if data_args.block_size is None:
         block_size = tokenizer.model_max_length
-        if block_size > 1024:
+        if block_size > config.max_position_embeddings:
             logger.warning(
-                "The chosen tokenizer supports a `model_max_length` that is longer than the default `block_size` value"
-                " of 1024. If you would like to use a longer `block_size` up to `tokenizer.model_max_length` you can"
-                " override this default with `--block_size xxx`."
+                f"The tokenizer picked seems to have a very large `model_max_length` ({tokenizer.model_max_length}). "
+                f"Using block_size={min(1024, config.max_position_embeddings)} instead. You can change that default value by passing --block_size xxx."
             )
-            block_size = 1024
+            block_size = min(1024, config.max_position_embeddings)
     else:
         if data_args.block_size > tokenizer.model_max_length:
             logger.warning(

examples/pytorch/language-modeling/run_clm_no_trainer.py

@@ -440,13 +440,12 @@ def main():
 
     if args.block_size is None:
         block_size = tokenizer.model_max_length
-        if block_size > 1024:
+        if block_size > config.max_position_embeddings:
             logger.warning(
-                "The chosen tokenizer supports a `model_max_length` that is longer than the default `block_size` value"
-                " of 1024. If you would like to use a longer `block_size` up to `tokenizer.model_max_length` you can"
-                " override this default with `--block_size xxx`."
+                f"The tokenizer picked seems to have a very large `model_max_length` ({tokenizer.model_max_length}). "
+                f"Using block_size={min(1024, config.max_position_embeddings)} instead. You can change that default value by passing --block_size xxx."
             )
-        block_size = 1024
+            block_size = min(1024, config.max_position_embeddings)
     else:
         if args.block_size > tokenizer.model_max_length:
             logger.warning(