Changes variable name to encoder_config

Signed-off-by: Flavia Beo <[email protected]>

tests/model_executor/test_model_load_with_params.py

@@ -29,8 +29,8 @@ def test_model_loading_with_params(vllm_runner):
         model_tokenizer = model.model.llm_engine.tokenizer
 
         # asserts on the bert model config file
-        assert model_config.bert_config["max_seq_length"] == 512
-        assert model_config.bert_config["do_lower_case"]
+        assert model_config.encoder_config["max_seq_length"] == 512
+        assert model_config.encoder_config["do_lower_case"]
 
         # asserts on the pooling config files
         assert model_config.pooler_config.pooling_type == PoolingType.CLS.name

tests/test_config.py

@@ -168,7 +168,7 @@ def test_get_bert_tokenization_sentence_transformer_config():
         revision=None,
     )
 
-    bert_bge_model_config = bge_model_config._get_bert_config()
+    bert_bge_model_config = bge_model_config._get_encoder_config()
 
     assert bert_bge_model_config["max_seq_length"] == 512
     assert bert_bge_model_config["do_lower_case"]

vllm/config.py

@@ -197,7 +197,7 @@ def __init__(
                                     code_revision, rope_scaling, rope_theta,
                                     config_format)
         self.hf_text_config = get_hf_text_config(self.hf_config)
-        self.bert_config = self._get_bert_config()
+        self.encoder_config = self._get_encoder_config()
         self.hf_image_processor_config = get_hf_image_processor_config(
             self.model, revision)
         self.dtype = _get_and_verify_dtype(self.hf_text_config, dtype)
@@ -231,7 +231,7 @@ def __init__(
             disable_sliding_window=self.disable_sliding_window,
             sliding_window_len=self.get_hf_config_sliding_window(),
             spec_target_max_model_len=spec_target_max_model_len,
-            bert_config=self.bert_config)
+            encoder_config=self.encoder_config)
         self.served_model_name = get_served_model_name(model,
                                                        served_model_name)
         self.multimodal_config = self._init_multimodal_config(
@@ -275,7 +275,7 @@ def _init_multimodal_config(
 
         return None
 
-    def _get_bert_config(self):
+    def _get_encoder_config(self):
         return get_sentence_transformer_tokenizer_config(
             self.model, self.revision)
 
@@ -1808,7 +1808,7 @@ def _get_and_verify_max_len(
     disable_sliding_window: bool,
     sliding_window_len: Optional[Union[int, List[Optional[int]]]],
     spec_target_max_model_len: Optional[int] = None,
-    bert_config: Optional[Any] = None,
+    encoder_config: Optional[Any] = None,
 ) -> int:
     """Get and verify the model's maximum length."""
     derived_max_model_len = float("inf")
@@ -1891,8 +1891,8 @@ def _get_and_verify_max_len(
                     "original_max_position_embeddings"]
             derived_max_model_len *= scaling_factor
 
-    if bert_config and "max_seq_length" in bert_config:
-        derived_max_model_len = bert_config["max_seq_length"]
+    if encoder_config and "max_seq_length" in encoder_config:
+        derived_max_model_len = encoder_config["max_seq_length"]
 
     # If the user specified a max length, make sure it is smaller than the
     # derived length from the HF model config.

vllm/transformers_utils/config.py

@@ -377,15 +377,15 @@ def get_sentence_transformer_tokenizer_config(model: str,
             "sentence_xlm-roberta_config.json",
             "sentence_xlnet_config.json",
     ]:
-        bert_dict = get_hf_file_to_dict(config_name, model, revision, token)
-        if bert_dict:
+        encoder_dict = get_hf_file_to_dict(config_name, model, revision, token)
+        if encoder_dict:
             break
 
-    if not bert_dict:
+    if not encoder_dict:
         return None
 
-    if all(k in bert_dict for k in ("max_seq_length", "do_lower_case")):
-        return bert_dict
+    if all(k in encoder_dict for k in ("max_seq_length", "do_lower_case")):
+        return encoder_dict
     return None
 
 

vllm/transformers_utils/tokenizer_group/__init__.py

@@ -25,9 +25,9 @@ def init_tokenizer_from_configs(model_config: ModelConfig,
                        trust_remote_code=model_config.trust_remote_code,
                        revision=model_config.tokenizer_revision)
 
-    if (model_config.bert_config is not None
-            and "do_lower_case" in model_config.bert_config):
-        init_kwargs["do_lower_case"] = model_config.bert_config[
+    if (model_config.encoder_config is not None
+            and "do_lower_case" in model_config.encoder_config):
+        init_kwargs["do_lower_case"] = model_config.encoder_config[
             "do_lower_case"]
 
     return get_tokenizer_group(parallel_config.tokenizer_pool_config,