Fixes bugs when using gpt series model (#3348)

* fixes minor bugs when using gpt series model

* pacify pre-commit

---------
Co-authored-by: Baber <baber@hey.com>

lm_eval/models/openai_completions.py

@@ -150,6 +150,8 @@ class LocalChatCompletion(LocalCompletionsAPI):
     def __init__(
         self,
         base_url=None,
+        tokenizer_backend=None,
+        tokenized_requests=None,
         verify_certificate=True,
         ca_cert_path=None,
         auth_token=None,
@@ -157,8 +159,8 @@ class LocalChatCompletion(LocalCompletionsAPI):
     ):
         super().__init__(
             base_url=base_url,
-            tokenizer_backend=None,
-            tokenized_requests=None,
+            tokenizer_backend=tokenizer_backend,
+            tokenized_requests=tokenized_requests,
             verify_certificate=verify_certificate,
             ca_cert_path=ca_cert_path,
             auth_token=auth_token,
@@ -335,9 +337,12 @@ class OpenAIChatCompletion(LocalChatCompletion):
             "seed": seed,
             **gen_kwargs,
         }
-        if "o1" in self.model or "5" in self.model:
+        if (
+            "o1" in self.model
+            or "5" in self.model
+            or "o3" in self.model
+            or "o4" in self.model
+        ):
             output.pop("stop")
             output["temperature"] = 1
-        elif "o3" in self.model:
-            output.pop("temperature")
         return output