Update modeling_reasonir_8b.py

modeling_reasonir_8b.py

@@ -1683,8 +1683,7 @@ class ReasonIRModel(LlamaModel):
         self.pooling_method = "mean"
         self.normalized = True
         self.embed_eos = ""
-        self.reasonir_config = config
-        self.tokenizer = AutoTokenizer.from_pretrained('reasonir/ReasonIR-8B')
+        self.tokenizer = AutoTokenizer.from_pretrained(config._name_or_path, padding_side='right', trust_remote_code=True)
         
     def encode_queries(self, queries: Union[List[str], str], **kwargs) -> np.ndarray:
         """Used for encoding the queries of retrieval or reranking tasks"""
@@ -1716,6 +1715,8 @@ class ReasonIRModel(LlamaModel):
         **kwargs,
     ) -> np.ndarray:
 
+        self.eval()
+
         # get number of gpus
         num_gpus = torch.cuda.device_count()
         if num_gpus > 0: