Merge pull request #179 from AIRobotZhang/patch-1

Update loader.py

Author: tnlin

IOPO/Method-IOPO/src/llamafactory/data/loader.py

@@ -140,14 +140,14 @@ def _get_merged_dataset(
     model_args: "ModelArguments",
     data_args: "DataArguments",
     training_args: "Seq2SeqTrainingArguments",
-    stage: Literal["pt", "sft", "rm", "ppo", "kto"],
+    stage: Literal["pt", "sft", "rm", "ppo", "kto", "iopo"],
 ) -> Optional[Union["Dataset", "IterableDataset"]]:
     if dataset_names is None:
         return None
 
     datasets = []
     for dataset_attr in get_dataset_list(dataset_names, data_args.dataset_dir):
-        if (stage == "rm" and dataset_attr.ranking is False) or (stage != "rm" and dataset_attr.ranking is True):
+        if ((stage == "rm" or stage == "iopo") and dataset_attr.ranking is False) or ((stage != "rm" and stage != "iopo") and dataset_attr.ranking is True):
             raise ValueError("The dataset is not applicable in the current training stage.")
 
         datasets.append(_load_single_dataset(dataset_attr, model_args, data_args, training_args))
@@ -199,7 +199,7 @@ def get_dataset(
     model_args: "ModelArguments",
     data_args: "DataArguments",
     training_args: "Seq2SeqTrainingArguments",
-    stage: Literal["pt", "sft", "rm", "ppo", "kto"],
+    stage: Literal["pt", "sft", "rm", "ppo", "kto", "iopo"],
     tokenizer: "PreTrainedTokenizer",
     processor: Optional["ProcessorMixin"] = None,
 ) -> "DatasetModule":