Merge pull request #1211 from hanhainebula/master

Fix bugs

Author: 545999961

FlagEmbedding/abc/finetune/embedder/AbsDataset.py

@@ -196,7 +196,7 @@ def __call__(self, features):
             d_collated = self.tokenizer.pad(
                 passages_inputs,
                 padding=self.padding,
-                max_length=self.query_max_len,
+                max_length=self.passage_max_len,
                 pad_to_multiple_of=self.pad_to_multiple_of,
                 return_tensors=self.return_tensors
             )
@@ -562,7 +562,7 @@ def __call__(self, features):
                 q_collated.append(self.tokenizer.pad(
                     sub_features,
                     padding=self.padding,
-                    max_length=self.passage_max_len,
+                    max_length=self.query_max_len,
                     pad_to_multiple_of=self.pad_to_multiple_of,
                     return_tensors=self.return_tensors,
                 ))

scripts/add_reranker_score.py

@@ -29,7 +29,7 @@ class ModelArgs:
         default=None, metadata={"help": "The reranker name or path."}
     )
     reranker_model_class: Optional[str] = field(
-        default=None, metadata={"help": "The reranker model class. Available classes: ['auto', 'encoder-only-base', 'decoder-only-base', 'decoder-only-layerwise', 'decoder-only-lightweight']. Default: auto.", "choices": ["encoder-only-base", "decoder-only-base", "decoder-only-layerwise", "decoder-only-lightweight"]}
+        default=None, metadata={"help": "The reranker model class. Available classes: ['encoder-only-base', 'decoder-only-base', 'decoder-only-layerwise', 'decoder-only-lightweight']. Default: None. For the custom model, you need to specify the model class.", "choices": ["encoder-only-base", "decoder-only-base", "decoder-only-layerwise", "decoder-only-lightweight"]}
     )
     reranker_peft_path: Optional[str] = field(
         default=None, metadata={"help": "The reranker peft path."}