move model type sync outside

src/leaderboard/read_evals.py

@@ -128,18 +128,6 @@ class EvalResult:
         except Exception as e:
             print(f"Could not find request file for {self.org}/{self.model} -- path: {requests_path} -- {e}")
 
-    def update_model_type_with_open_llm_request_file(self, open_llm_requests_path):
-        """Finds the relevant request file for the current model and updates info with it"""
-        request_file = get_request_file_for_model_open_llm(open_llm_requests_path, self.full_model, self.precision.value.name)
-
-        if request_file:
-            try:
-                with open(request_file, "r") as f:
-                    request = json.load(f)
-                self.model_type = ModelType.from_str(request.get("model_type", "Unknown"))
-            except Exception as e:
-                pass
-
     def is_complete(self) -> bool:
         for task in Tasks:
             if task.value.benchmark not in self.results:
@@ -216,10 +204,23 @@ def get_request_file_for_model_open_llm(requests_path, model_name, precision):
                 request_file = tmp_request_file
     return request_file
 
+def update_model_type_with_open_llm_request_file(result, open_llm_requests_path):
+    """Finds the relevant request file for the current model and updates info with it"""
+    request_file = get_request_file_for_model_open_llm(open_llm_requests_path, result.full_model, result.precision.value.name)
+
+    if request_file:
+        try:
+            with open(request_file, "r") as f:
+                request = json.load(f)
+            open_llm_model_type = request.get("model_type", "Unknown")
+            if open_llm_model_type != "Unknown":
+                result.model_type = ModelType.from_str(open_llm_model_type)
+        except Exception as e:
+            pass
+    return result
 
 def get_raw_eval_results(results_path: str,
                          requests_path: str,
-                         requests_path_open_llm: Optional[str] = None,
                          is_backend: bool = False) -> list[EvalResult]:
     """From the path of the results folder root, extract all needed info for results"""
     model_result_filepaths = []
@@ -243,8 +244,6 @@ def get_raw_eval_results(results_path: str,
         # Creation of result
         eval_result = EvalResult.init_from_json_file(model_result_filepath, is_backend=is_backend)
         eval_result.update_with_request_file(requests_path)
-        if requests_path_open_llm is not None:
-            eval_result.update_model_type_with_open_llm_request_file(requests_path_open_llm)
         # Store results of same eval together
         eval_name = eval_result.eval_name
         if eval_name in eval_results.keys():

src/populate.py

@@ -1,13 +1,13 @@
 import json
 import os
-
+from tqdm import tqdm
 import copy
 import pandas as pd
 
 from src.display.formatting import has_no_nan_values, make_clickable_model
 from src.display.utils import AutoEvalColumn, EvalQueueColumn
 from src.leaderboard.filter_models import filter_models
-from src.leaderboard.read_evals import get_raw_eval_results, EvalResult
+from src.leaderboard.read_evals import get_raw_eval_results, EvalResult, update_model_type_with_open_llm_request_file
 
 from src.backend.envs import Tasks as BackendTasks
 from src.display.utils import Tasks
@@ -21,6 +21,9 @@ def get_leaderboard_df(results_path: str,
                        is_backend: bool = False) -> tuple[list[EvalResult], pd.DataFrame]:
     # Returns a list of EvalResult
     raw_data: list[EvalResult] = get_raw_eval_results(results_path, requests_path, requests_path_open_llm)
+    for result_idx in tqdm(range(len(raw_data)), desc="updating model type with open llm leaderboard"):
+        if requests_path_open_llm:
+            raw_data[result_idx] = update_model_type_with_open_llm_request_file(raw_data[result_idx], requests_path_open_llm)
 
     all_data_json_ = [v.to_dict() for v in raw_data if v.is_complete()]