Modified dataframe to adapt to new format with real data. Fixed html code

app.py

@@ -18,21 +18,24 @@ st.set_page_config(page_title="IVACE Leaderboard", layout="wide")
 request_file = Path("user_request/") / f"data_{uuid.uuid4()}.json"
 request_folder = request_file.parent
 
-# column order
-columns = [
-    "eval_name",
-    "Model",
-    "Type",
-    "Average ⬆️",
-    "IFEval",
-    "MMLU-PRO",
-    "GPQA",
-    "MUSR",
-    "CO₂ cost (kg)",
-]
 
+# columns = [
+#     "eval_name",
+#     "Model",
+#     "Type",
+#     "Average ⬆️",
+#     "IFEval",
+#     "MMLU-PRO",
+#     "GPQA",
+#     "MUSR",
+#     "CO₂ cost (kg)",
+# ]
 # languages
-lang_list = ["Spanish", "Galician", "Basque", "Argentinian", "Chilean"]
+lang_list = ["Spain", "Costa Rica", "Mexico", "Peru", "Uruguay"]
+
+# column order
+model_columns = ["model_name", "url", "type"]
+task_columns = [f"tass_{lang.lower().replace(' ', '_')}" for lang in lang_list]
 
 scheduler = CommitScheduler(
     repo_id="iberbench/ivace-user-request",
@@ -55,33 +58,43 @@ def log_submission(input_dict: dict) -> None:
             f.write("\n")
 
 
-def get_url(html_content: str) -> str:
-    match = re.search(r'href=["\'](https?://[^\s"\']+)', html_content)
-    if match:
-        url = match.group(1)
-        return url
-    else:
-        raise ValueError("Url not found in the link")
+# def get_url(html_content: str) -> str:
+#     match = re.search(r'href=["\'](https?://[^\s"\']+)', html_content)
+#     if match:
+#         url = match.group(1)
+#         return url
+#     else:
+#         raise ValueError("Url not found in the link")
+
+
+def get_lang_columns(columns: list, lang: str):
+    """Filter columns per language"""
+    lang_norm = lang.lower().replace(" ", "_")
+
+    return [col for col in columns if lang_norm in col]
 
 
 @st.cache_data
-def load_data() -> pd.DataFrame:
+def load_data(lang) -> pd.DataFrame:
     try:
         data = (
-            load_dataset("open-llm-leaderboard/contents")["train"]
+            load_dataset("iberbench/lm-eval-results-ac", token=st.secrets["HF_TOKEN"])["train"]
             .to_pandas()
-            .head(10)
         )
-        data = data[columns]
-        data["Model"] = data["Model"].apply(get_url)
-        data.sort_values(by="Average ⬆️", ascending=False, inplace=True)
-        data.reset_index(drop=True, inplace=True)
+        # filter lang columns
+        task_lang_columns = get_lang_columns(task_columns, lang)
+        data = data[model_columns + task_lang_columns]
+
+        # data["Model"] = data["Model"].apply(get_url)
+        # data.sort_values(by="Average ⬆️", ascending=False, inplace=True)
+        # data.reset_index(drop=True, inplace=True)
+
         # add column to apply filtering
         data["Active"] = False
 
         return data
     except FileNotFoundError:
-        st.error("open-llm-leaderboard/contents was not found in the hub")
+        st.error("iberbench/lm-eval-results-ac was not found in the hub")
         return pd.DataFrame()
 
 
@@ -129,12 +142,17 @@ def create_search_per_language(lang: str, search_dict: dict):
                 "Active",
             ] = True
 
+        # select columns to display
+        task_lang_columns = get_lang_columns(task_columns, lang)
+        columns = model_columns + task_lang_columns
+
         edited_data = st.data_editor(
             active_data(lang),
             column_order=columns,
             key=f"edited_data_{lang}",
             hide_index=False,
-            column_config={"Model": st.column_config.LinkColumn("Model")},
+            # column_config={"Model": st.column_config.LinkColumn("Model")},
+            column_config={"url": st.column_config.LinkColumn("url")},
         )
     else:
         st.write("No data found to display on leaderboard.")
@@ -142,8 +160,8 @@ def create_search_per_language(lang: str, search_dict: dict):
 
 # streamlit UI
 for lang in lang_list:
-    # todo: load a different dataset per language
-    leaderboard_data = load_data()
+    # todo: load a different dataset per language of load different column per lang
+    leaderboard_data = load_data(lang)
     if f"leaderboard_data_{lang}" not in st.session_state:
         st.session_state[f"leaderboard_data_{lang}"] = leaderboard_data
 

assets/html/02_technical_detail.html

@@ -5,11 +5,12 @@
   </p>
   <div style="background-color: #f5f5f5; padding: 1rem; border-radius: 5px; font-family: monospace; color: #212529;">
     <pre style="margin: 0; padding: 0; font-size: 1rem; white-space: pre-wrap; word-wrap: break-word;">
+      <code>
 from transformers import AutoConfig, AutoModel, AutoTokenizer
-
 config = AutoConfig.from_pretrained("your-username/your-model", revision="main")
 model = AutoModel.from_pretrained("your-username/your-model", revision="main")
 tokenizer = AutoTokenizer.from_pretrained("your-username/your-model", revision="main")
+      </code>
     </pre>
   </div>
   <a href="https://huggingface.co/docs/transformers/installation" target="_blank" style="color: #007BFF; text-decoration: underline; font-family: monospace;">Transformers documentation →</a>