Remove weight type

src/leaderboard/read_evals.py

@@ -1,37 +1,33 @@
-print("--- CONFIRMED: Running the modified version of read_evals.py ---")
-
-import glob
-import json
-import math
-import os
-from dataclasses import dataclass
-
-import dateutil
-import numpy as np
-
-from src.display.formatting import make_clickable_model
-from src.display.utils import AutoEvalColumn, ModelType, Tasks, Precision, WeightType
 from src.submission.check_validity import is_model_on_hub
+from src.display.utils import AutoEvalColumn, ModelType, Tasks, Precision
+from src.display.formatting import make_clickable_model
+import numpy as np
+import dateutil
+from dataclasses import dataclass
+import os
+import math
+import json
+import glob
+print("--- CONFIRMED: Running the modified version of read_evals.py ---")
 
 
 @dataclass
 class EvalResult:
     """Represents one full evaluation. Built from a combination of the result and request file for a given run.
     """
-    eval_name: str # org_model_precision (uid)
-    full_model: str # org/model (path on hub)
-    org: str 
+    eval_name: str  # org_model_precision (uid)
+    full_model: str  # org/model (path on hub)
+    org: str
     model: str
-    revision: str # commit hash, "" if main
+    revision: str  # commit hash, "" if main
     results: dict
     precision: Precision = Precision.Unknown
-    model_type: ModelType = ModelType.Unknown # Pretrained, fine tuned, ...
-    weight_type: WeightType = WeightType.Original # Original or Adapter
-    architecture: str = "Unknown" 
+    model_type: ModelType = ModelType.Unknown  # Pretrained, fine tuned, ...
+    architecture: str = "Unknown"
     license: str = "?"
     likes: int = 0
     num_params: int = 0
-    date: str = "" # submission date of request file
+    date: str = ""  # submission date of request file
     still_on_hub: bool = False
 
     @classmethod
@@ -46,7 +42,8 @@ class EvalResult:
         precision = Precision.from_str(config.get("model_dtype"))
 
         # Get model and org
-        org_and_model = config.get("model_name", config.get("model_args", None))
+        org_and_model = config.get(
+            "model_name", config.get("model_args", None))
         org_and_model = org_and_model.split("/", 1)
 
         if len(org_and_model) == 1:
@@ -74,7 +71,8 @@ class EvalResult:
             task = task.value
 
             # We average all scores of a given metric (not all metrics are present in all files)
-            accs = np.array([v.get(task.metric, None) for k, v in data["results"].items() if task.benchmark == k])
+            accs = np.array([v.get(task.metric, None)
+                            for k, v in data["results"].items() if task.benchmark == k])
             if accs.size == 0 or any([acc is None for acc in accs]):
                 continue
 
@@ -87,37 +85,38 @@ class EvalResult:
             org=org,
             model=model,
             results=results,
-            precision=precision,  
-            revision= config.get("model_sha", ""),
+            precision=precision,
+            revision=config.get("model_sha", ""),
             still_on_hub=still_on_hub,
             architecture=architecture
         )
 
     def update_with_request_file(self, requests_path):
         """Finds the relevant request file for the current model and updates info with it"""
-        request_file = get_request_file_for_model(requests_path, self.full_model, self.precision.value.name)
+        request_file = get_request_file_for_model(
+            requests_path, self.full_model, self.precision.value.name)
 
         try:
             with open(request_file, "r") as f:
                 request = json.load(f)
             self.model_type = ModelType.from_str(request.get("model_type", ""))
-            self.weight_type = WeightType[request.get("weight_type", "Original")]
             self.license = request.get("license", "?")
             self.likes = request.get("likes", 0)
             self.num_params = request.get("params", 0)
             self.date = request.get("submitted_time", "")
         except Exception:
-            print(f"Could not find request file for {self.org}/{self.model} with precision {self.precision.value.name}")
+            print(
+                f"Could not find request file for {self.org}/{self.model} with precision {self.precision.value.name}")
 
     def to_dict(self):
         """Converts the Eval Result to a dict compatible with our dataframe display"""
-        average = sum([v for v in self.results.values() if v is not None]) / len(Tasks)
+        average = sum([v for v in self.results.values()
+                      if v is not None]) / len(Tasks)
         data_dict = {
             "eval_name": self.eval_name,  # not a column, just a save name,
             AutoEvalColumn.precision.name: self.precision.value.name,
             AutoEvalColumn.model_type.name: self.model_type.value.name,
             AutoEvalColumn.model_type_symbol.name: self.model_type.value.symbol,
-            AutoEvalColumn.weight_type.name: self.weight_type.value.name,
             AutoEvalColumn.architecture.name: self.architecture,
             AutoEvalColumn.model.name: make_clickable_model(self.full_model),
             AutoEvalColumn.revision.name: self.revision,
@@ -167,7 +166,8 @@ def get_raw_eval_results(results_path: str, requests_path: str) -> list[EvalResu
 
         # Sort the files by date
         try:
-            files.sort(key=lambda x: x.removesuffix(".json").removeprefix("results_")[:-7])
+            files.sort(key=lambda x: x.removesuffix(
+                ".json").removeprefix("results_")[:-7])
         except dateutil.parser._parser.ParserError:
             files = [files[-1]]
 
@@ -183,14 +183,15 @@ def get_raw_eval_results(results_path: str, requests_path: str) -> list[EvalResu
         # Store results of same eval together
         eval_name = eval_result.eval_name
         if eval_name in eval_results.keys():
-            eval_results[eval_name].results.update({k: v for k, v in eval_result.results.items() if v is not None})
+            eval_results[eval_name].results.update(
+                {k: v for k, v in eval_result.results.items() if v is not None})
         else:
             eval_results[eval_name] = eval_result
 
     results = []
     for v in eval_results.values():
         try:
-            v.to_dict() # we test if the dict version is complete
+            v.to_dict()  # we test if the dict version is complete
             results.append(v)
         except Exception as e:
             print(f"--- DEBUG: SKIPPING RESULT FILE. ERROR IS: ---")