cleanup + refactor

app.py

@@ -9,9 +9,9 @@ from utils import (
     PROPRIETARY_LICENSES,
     CAT_NAME_TO_EXPLANATION,
     download_latest_data_from_space,
+    get_constants,
 )
 
-# with gr.NO_RELOAD:
 ###################
 ### Load Data
 ###################
@@ -72,43 +72,21 @@ merged_dfs = {k: format_data(v) for k, v in merged_dfs.items()}
 
 
 # get constants
-filter_ranges = {}
-for k, df in merged_dfs.items():
-    filter_ranges[k] = {
-        "min_elo_score": df["rating"].min().round(),
-        "max_elo_score": df["rating"].max().round(),
-        "upper_models_per_month": int(
-            df.groupby(["Month-Year", "License"])["rating"]
-            .apply(lambda x: x.count())
-            .max()
-        ),
-    }
-
-min_elo_score = float("inf")
-max_elo_score = float("-inf")
-upper_models_per_month = 0
-
-for key, value in filter_ranges.items():
-    min_elo_score = min(min_elo_score, value["min_elo_score"])
-    max_elo_score = max(max_elo_score, value["max_elo_score"])
-    upper_models_per_month = max(
-        upper_models_per_month, value["upper_models_per_month"]
-    )
-
+min_elo_score, max_elo_score, upper_models_per_month = get_constants(merged_dfs)
 
 date_updated = elo_results["full"]["last_updated_datetime"].split(" ")[0]
 
 
-def get_data_split(dfs, set_name):
-    df = dfs[set_name].copy(deep=True)
-    return df.reset_index(drop=True)
-
-
 ###################
 ### Plot Data
 ###################
 
 
+def get_data_split(dfs, set_name):
+    df = dfs[set_name].copy(deep=True)
+    return df.reset_index(drop=True)
+
+
 def build_plot(min_score, max_models_per_month, toggle_annotations, set_selector):
 
     df = get_data_split(merged_dfs, set_name=set_selector)
@@ -172,7 +150,7 @@ with gr.Blocks(
     gr.Markdown(
         """
         <div style="text-align: center; max-width: 650px; margin: auto;">
-            <h1 style="font-weight: 900; margin-top: 5px;">🔬 Progress Tracker: Proprietary vs Open LLMs
+            <h1 style="font-weight: 900; margin-top: 5px;">🔬 Progress Tracker: Open vs. Proprietary LLMs
             </h1>
             <p style="text-align: left; margin-top: 10px; margin-bottom: 10px; line-height: 20px;">
             This app visualizes the progress of proprietary and open-source LLMs in the LMSYS Arena ELO leaderboard. The idea is inspired by <a href="https://www.linkedin.com/posts/maxime-labonne_arena-elo-graph-updated-with-new-models-activity-7187062633735368705-u2jB?utm_source=share&utm_medium=member_desktop">this great work</a> from <a href="https://huggingface.co/mlabonne/">Maxime Labonne</a>.

utils.py

@@ -58,3 +58,40 @@ def download_latest_data_from_space(
         repo_type="space",
     )
     return latest_filepath_local
+
+
+def get_constants(dfs):
+    """
+    Calculate and return the minimum and maximum Elo scores, as well as the maximum number of models per month.
+
+    Parameters:
+    - dfs (dict): A dictionary containing DataFrames for different categories.
+
+    Returns:
+    - min_elo_score (float): The minimum Elo score across all DataFrames.
+    - max_elo_score (float): The maximum Elo score across all DataFrames.
+    - upper_models_per_month (int): The maximum number of models per month per license across all DataFrames.
+    """
+    filter_ranges = {}
+    for k, df in dfs.items():
+        filter_ranges[k] = {
+            "min_elo_score": df["rating"].min().round(),
+            "max_elo_score": df["rating"].max().round(),
+            "upper_models_per_month": int(
+                df.groupby(["Month-Year", "License"])["rating"]
+                .apply(lambda x: x.count())
+                .max()
+            ),
+        }
+
+    min_elo_score = float("inf")
+    max_elo_score = float("-inf")
+    upper_models_per_month = 0
+
+    for _, value in filter_ranges.items():
+        min_elo_score = min(min_elo_score, value["min_elo_score"])
+        max_elo_score = max(max_elo_score, value["max_elo_score"])
+        upper_models_per_month = max(
+            upper_models_per_month, value["upper_models_per_month"]
+        )
+    return min_elo_score, max_elo_score, upper_models_per_month