[BFCL] Using model_name attribute (#1195)

Standardize and make use of the `model_name` attribute in
`model_config.py` to indicate the name of the model as used in the
vendor API or on Hugging Face (may not be unique). This avoids any
further processing within each handler (eg,
`self.model_name.replace("-FC", "")`).

---------

Co-authored-by: Huanzhi Mao <[email protected]>

Author: ranashreyas

berkeley-function-call-leaderboard/bfcl_eval/.env.example

@@ -3,6 +3,9 @@ SERPAPI_API_KEY=
 
 # Provide the API key for the model(s) you intend to use
 OPENAI_API_KEY=sk-XXXXXX
+OPENAI_DEFAULT_HEADERS=
+OPENAI_BASE_URL=
+
 ANTHROPIC_API_KEY=
 # We use Google AI Studio to inference Google Gemini models
 GOOGLE_API_KEY=

berkeley-function-call-leaderboard/bfcl_eval/_llm_response_generation.py

@@ -64,9 +64,12 @@ def get_args():
 
 def build_handler(model_name, temperature):
     config = MODEL_CONFIG_MAPPING[model_name]
-    handler = config.model_handler(model_name, temperature)
-    # Propagate config flags to the handler instance
-    handler.is_fc_model = config.is_fc_model
+    handler = config.model_handler(
+        model_name=config.model_name,
+        temperature=temperature,
+        registry_name=model_name,
+        is_fc_model=config.is_fc_model,
+    )
     return handler
 
 
@@ -206,7 +209,11 @@ def generate_results(args, model_name, test_cases_total):
         is_oss_model = True
         # For OSS models, if the user didn't explicitly set the number of threads,
         # we default to 100 threads to speed up the inference.
-        num_threads = args.num_threads if args.num_threads is not None else LOCAL_SERVER_MAX_CONCURRENT_REQUEST
+        num_threads = (
+            args.num_threads
+            if args.num_threads is not None
+            else LOCAL_SERVER_MAX_CONCURRENT_REQUEST
+        )
     else:
         handler: BaseHandler
         is_oss_model = False