diff --git a/unsloth/models/rl.py b/unsloth/models/rl.py index 24cd5e60af..6da8bf4604 100644 --- a/unsloth/models/rl.py +++ b/unsloth/models/rl.py @@ -358,7 +358,7 @@ def _patch_trl_rl_trainers(trainer_file = "grpo_trainer"): "from unsloth_zoo.vision_utils import UnslothVisionDataCollator\n"\ "if not isinstance(data_collator, UnslothVisionDataCollator):\n"\ " if isinstance(data_collator, DataCollatorForSeq2Seq) and 'labels' not in train_dataset.column_names:\n"\ - " data_collator = TransformersDataCollatorForLanguageModeling(__tokenizer, mlm = False)\n"\ + " data_collator = TransformersDataCollatorForLanguageModeling(__tokenizer, mlm = False, mlm_probability = 0.0)\n"\ " elif isinstance(data_collator, TransformersDataCollatorForLanguageModeling) and 'labels' in train_dataset.column_names:\n"\ " data_collator = DataCollatorForSeq2Seq(__tokenizer)\n"\ "else:\n"\