Upload app.py (#7)

- Upload app.py (d99d0376d489de9ff899d60ce46d59ec04bc4621)


Co-authored-by: CultriX <CultriX@users.noreply.huggingface.co>

app.py

@@ -10,6 +10,7 @@ from huggingface_hub.utils import RepositoryNotFoundError, RevisionNotFoundError
 from yall import create_yall
 
 
+
 def convert_markdown_table_to_dataframe(md_content):
     """
     Converts markdown table to Pandas DataFrame, handling special characters and links,
@@ -36,10 +37,10 @@ def convert_markdown_table_to_dataframe(md_content):
 
     return df
 
-@st.cache_data 
+@st.cache_data
 def get_model_info(df):
     api = HfApi()
-    
+
     # Initialize new columns for likes and tags
     df['Likes'] = None
     df['Tags'] = None
@@ -58,7 +59,8 @@ def get_model_info(df):
 
     return df
 
-    
+
+
 def create_bar_chart(df, category):
     """Create and display a bar chart for a given category."""
     st.write(f"### {category} Scores")
@@ -66,12 +68,12 @@ def create_bar_chart(df, category):
     # Sort the DataFrame based on the category score
     sorted_df = df[['Model', category]].sort_values(by=category, ascending=True)
 
-    # Create the bar chart with color gradient
+    # Create the bar chart with a color gradient (using 'Viridis' color scale as an example)
     fig = go.Figure(go.Bar(
         x=sorted_df[category],
         y=sorted_df['Model'],
         orientation='h',
-        marker=dict(color=sorted_df[category], colorscale='Inferno')
+        marker=dict(color=sorted_df[category], colorscale='Twilight')  # You can change 'Viridis' to another color scale
     ))
 
     # Update layout for better readability
@@ -79,15 +81,18 @@ def create_bar_chart(df, category):
         margin=dict(l=20, r=20, t=20, b=20)
     )
 
-    st.plotly_chart(fig, use_container_width=True, height=len(df)*35)
+    # Adjust the height of the chart based on the number of rows in the DataFrame
+    st.plotly_chart(fig, use_container_width=True, height=len(df) * 35)
+
+# Example usage:
+# create_bar_chart(your_dataframe, 'Your_Category')
+
 
-    
 def main():
     st.set_page_config(page_title="YALL - Yet Another LLM Leaderboard", layout="wide")
 
     st.title("🏆 YALL - Yet Another LLM Leaderboard")
-    st.markdown("Leaderboard made with 🧐 [LLM AutoEval](https://github.com/mlabonne/llm-autoeval) using [Nous](https://huggingface.co/NousResearch) benchmark suite.")
-
+    st.markdown("Leaderboard made with 🧐 [LLM AutoEval](https:
     content = create_yall()
     tab1, tab2 = st.tabs(["🏆 Leaderboard", "📝 About"])
 
@@ -96,7 +101,7 @@ def main():
         if content:
             try:
                 score_columns = ['Average', 'AGIEval', 'GPT4All', 'TruthfulQA', 'Bigbench']
-    
+
                 # Display dataframe
                 full_df = convert_markdown_table_to_dataframe(content)
                 for col in score_columns:
@@ -105,18 +110,15 @@ def main():
                 full_df = get_model_info(full_df)
                 full_df['Tags'] = full_df['Tags'].fillna('')
                 df = pd.DataFrame(columns=full_df.columns)
-                
-                # Toggles
-                col1, col2, col3 = st.columns(3)
-                with col1:
-                    show_phi = st.checkbox("Phi (2.8B)", value=True)
-                with col2:
-                    show_mistral = st.checkbox("Mistral (7B)", value=True)
-                with col3:
-                    show_other = st.checkbox("Other", value=True)
 
+                # Toggles for filtering by tags
+                show_phi = st.checkbox("Phi (2.8B)", value=True)
+                show_mistral = st.checkbox("Mistral (7B)", value=True)
+                show_other = st.checkbox("Other", value=True)
+
+                # Create a DataFrame based on selected filters
                 dfs_to_concat = []
-                
+
                 if show_phi:
                     dfs_to_concat.append(full_df[full_df['Tags'].str.lower().str.contains('phi,|phi-msft,')])
                 if show_mistral:
@@ -124,15 +126,19 @@ def main():
                 if show_other:
                     other_df = full_df[~full_df['Tags'].str.lower().str.contains('phi,|phi-msft,|mistral,')]
                     dfs_to_concat.append(other_df)
-                
+
                 # Concatenate the DataFrames
                 if dfs_to_concat:
                     df = pd.concat(dfs_to_concat, ignore_index=True)
-                
-                # Sort values
-                df = df.sort_values(by='Average', ascending=False)
-                
-                # Display the DataFrame
+
+                # Add a search bar
+                search_query = st.text_input("Search models", "")
+
+                # Filter the DataFrame based on the search query
+                if search_query:
+                    df = df[df['Model'].str.contains(search_query, case=False)]
+
+                # Display the filtered DataFrame or the entire leaderboard
                 st.dataframe(
                     df[['Model'] + score_columns + ['Likes', 'URL']],
                     use_container_width=True,
@@ -145,26 +151,41 @@ def main():
                         "URL": st.column_config.LinkColumn("URL"),
                     },
                     hide_index=True,
-                    height=len(df)*37,
+                    height=len(df) * 37,
                 )
 
+                # Add a button to export data to CSV
+                if st.button("Export to CSV"):
+                    # Export the DataFrame to CSV
+                    csv_data = df.to_csv(index=False)
+
+                    # Create a link to download the CSV file
+                    st.download_button(
+                        label="Download CSV",
+                        data=csv_data,
+                        file_name="leaderboard.csv",
+                        key="download-csv",
+                        help="Click to download the CSV file",
+                    )
+
                 # Full-width plot for the first category
                 create_bar_chart(df, score_columns[0])
-    
+
                 # Next two plots in two columns
                 col1, col2 = st.columns(2)
                 with col1:
                     create_bar_chart(df, score_columns[1])
                 with col2:
                     create_bar_chart(df, score_columns[2])
-    
+
                 # Last two plots in two columns
                 col3, col4 = st.columns(2)
                 with col3:
                     create_bar_chart(df, score_columns[3])
                 with col4:
                     create_bar_chart(df, score_columns[4])
-    
+
+
             except Exception as e:
                 st.error("An error occurred while processing the markdown table.")
                 st.error(str(e))
@@ -176,26 +197,18 @@ def main():
         st.markdown('''
             ### Nous benchmark suite
 
-            Popularized by [Teknium](https://huggingface.co/teknium) and [NousResearch](https://huggingface.co/NousResearch), this benchmark suite aggregates four benchmarks:
-            
-            * [**AGIEval**](https://arxiv.org/abs/2304.06364) (0-shot): `agieval_aqua_rat,agieval_logiqa_en,agieval_lsat_ar,agieval_lsat_lr,agieval_lsat_rc,agieval_sat_en,agieval_sat_en_without_passage,agieval_sat_math`
-            * **GPT4ALL** (0-shot): `hellaswag,openbookqa,winogrande,arc_easy,arc_challenge,boolq,piqa`
-            * [**TruthfulQA**](https://arxiv.org/abs/2109.07958) (0-shot): `truthfulqa_mc`
-            * [**Bigbench**](https://arxiv.org/abs/2206.04615) (0-shot): `bigbench_causal_judgement,bigbench_date_understanding,bigbench_disambiguation_qa,bigbench_geometric_shapes,bigbench_logical_deduction_five_objects,bigbench_logical_deduction_seven_objects,bigbench_logical_deduction_three_objects,bigbench_movie_recommendation,bigbench_navigate,bigbench_reasoning_about_colored_objects,bigbench_ruin_names,bigbench_salient_translation_error_detection,bigbench_snarks,bigbench_sports_understanding,bigbench_temporal_sequences,bigbench_tracking_shuffled_objects_five_objects,bigbench_tracking_shuffled_objects_seven_objects,bigbench_tracking_shuffled_objects_three_objects`
-
+            Popularized by [Teknium](https:
+            * [**AGIEval**](https:            * **GPT4ALL** (0-shot): `hellaswag,openbookqa,winogrande,arc_easy,arc_challenge,boolq,piqa`
+            * [**TruthfulQA**](https:            * [**Bigbench**](https:
             ### Reproducibility
 
-            You can easily reproduce these results using 🧐 [LLM AutoEval](https://github.com/mlabonne/llm-autoeval/tree/master), a colab notebook that automates the evaluation process (benchmark: `nous`). This will upload the results to GitHub as gists. You can find the entire table with the links to the detailed results [here](https://gist.github.com/mlabonne/90294929a2dbcb8877f9696f28105fdf).
-
+            You can easily reproduce these results using 🧐 [LLM AutoEval](https:
             ### Clone this space
 
             You can create your own leaderboard with your LLM AutoEval results on GitHub Gist. You just need to clone this space and specify two variables:
 
-            * Change the `gist_id` in [yall.py](https://huggingface.co/spaces/mlabonne/Yet_Another_LLM_Leaderboard/blob/main/yall.py#L126).
-            * Create "New Secret" in Settings > Variables and secrets (name: "github", value: [your GitHub token](https://github.com/settings/tokens))
-
-            A special thanks to [gblazex](https://huggingface.co/gblazex) for providing many evaluations.
-        ''')
+            * Change the `gist_id` in [yall.py](https:            * Create "New Secret" in Settings > Variables and secrets (name: "github", value: [your GitHub token](https:
+            A special thanks to [gblazex](https:        ''')
 
 if __name__ == "__main__":
-    main()
+    main()