Update app.py

app.py

@@ -5,69 +5,83 @@ import pandas as pd
 df_benchmark = pd.read_csv('leaderboard_data.csv')
 df_server = pd.read_csv('server_specs.csv')
 
-# Remove duplicated columns
-duplicate_cols = set(df_benchmark.columns) & set(df_server.columns)
-df_server_filtered = df_server.drop(columns=duplicate_cols)
+# Normalize column names (remove extra spaces, make case consistent)
+df_benchmark.columns = df_benchmark.columns.str.strip()
+df_server.columns = df_server.columns.str.strip()
 
-# Merge by "Server Name"
-df = pd.merge(df_benchmark, df_server_filtered, on="Server Name", how="left")
+# Debug output (optional)
+# st.write("Benchmark columns:", df_benchmark.columns.tolist())
+# st.write("Server columns:", df_server.columns.tolist())
 
-# Title and description
-st.title("MSNP Leaderboard")
-st.markdown("""
-[GitHub Repository](https://github.com/EvilFreelancer/llm-msnp-tests)  
-This leaderboard shows the performance of quantized GGUF models (via Ollama) on various CPU and GPU combinations in a single-node setup.
-""")
+# Ensure "Server Name" exists in both
+if "Server Name" not in df_benchmark.columns:
+    st.error('"Server Name" column not found in leaderboard_data.csv')
+elif "Server Name" not in df_server.columns:
+    st.error('"Server Name" column not found in server_specs.csv')
+else:
+    # Remove duplicated columns except "Server Name"
+    duplicate_cols = set(df_benchmark.columns) & set(df_server.columns) - {"Server Name"}
+    df_server_filtered = df_server.drop(columns=duplicate_cols)
 
-# Sidebar filters
-with st.sidebar:
-    st.header("Filters")
-    model_options = st.multiselect("Select model:", options=sorted(df["Model"].dropna().unique()))
-    gpu_options = st.multiselect("Select GPU:", options=sorted(df["GPUs"].dropna().unique()))
-    quantization_options = st.multiselect("Quantization:", options=sorted(df["Quantization"].dropna().unique()))
-    context_length_options = st.multiselect("Context length:", options=sorted(df["Context Length"].dropna().unique()))
-    start_type_options = st.multiselect("Start type:", options=sorted(df["Start Type"].dropna().unique()))
-    power_options = st.multiselect("Max GPU power (W):", options=sorted(df["GPU Max Power (W)"].dropna().unique()))
-    server_options = st.multiselect("Select server:", options=sorted(df["Server Name"].dropna().unique()))
+    # Merge datasets
+    df = pd.merge(df_benchmark, df_server_filtered, on="Server Name", how="left")
 
-# Apply filters
-filtered_df = df.copy()
-if model_options:
-    filtered_df = filtered_df[filtered_df["Model"].isin(model_options)]
-if gpu_options:
-    filtered_df = filtered_df[filtered_df["GPUs"].isin(gpu_options)]
-if quantization_options:
-    filtered_df = filtered_df[filtered_df["Quantization"].isin(quantization_options)]
-if context_length_options:
-    filtered_df = filtered_df[filtered_df["Context Length"].isin(context_length_options)]
-if start_type_options:
-    filtered_df = filtered_df[filtered_df["Start Type"].isin(start_type_options)]
-if power_options:
-    filtered_df = filtered_df[filtered_df["GPU Max Power (W)"].isin(power_options)]
-if server_options:
-    filtered_df = filtered_df[filtered_df["Server Name"].isin(server_options)]
+    # Continue with the rest of the app...
+    st.title("MSNP Leaderboard")
+    st.markdown("""
+    [GitHub Repository](https://github.com/EvilFreelancer/llm-msnp-tests)  
+    This leaderboard shows the performance of quantized GGUF models (via Ollama) on various CPU and GPU combinations in a single-node setup.
+    """)
 
-# Display table
-st.dataframe(filtered_df.sort_values(by="Tokens/sec", ascending=False).reset_index(drop=True))
+    # Filters
+    with st.sidebar:
+        st.header("Filters")
+        model_options = st.multiselect("Select model:", options=sorted(df["Model"].dropna().unique()))
+        gpu_options = st.multiselect("Select GPU:", options=sorted(df["GPUs"].dropna().unique()))
+        quantization_options = st.multiselect("Quantization:", options=sorted(df["Quantization"].dropna().unique()))
+        context_length_options = st.multiselect("Context length:", options=sorted(df["Context Length"].dropna().unique()))
+        start_type_options = st.multiselect("Start type:", options=sorted(df["Start Type"].dropna().unique()))
+        power_options = st.multiselect("Max GPU power (W):", options=sorted(df["GPU Max Power (W)"].dropna().unique()))
+        server_options = st.multiselect("Select server:", options=sorted(df["Server Name"].dropna().unique()))
 
-# Notes section
-st.markdown("## Notes")
-st.markdown("Click links in the 'Test Link' column for more details.")
+    # Apply filters
+    filtered_df = df.copy()
+    if model_options:
+        filtered_df = filtered_df[filtered_df["Model"].isin(model_options)]
+    if gpu_options:
+        filtered_df = filtered_df[filtered_df["GPUs"].isin(gpu_options)]
+    if quantization_options:
+        filtered_df = filtered_df[filtered_df["Quantization"].isin(quantization_options)]
+    if context_length_options:
+        filtered_df = filtered_df[filtered_df["Context Length"].isin(context_length_options)]
+    if start_type_options:
+        filtered_df = filtered_df[filtered_df["Start Type"].isin(start_type_options)]
+    if power_options:
+        filtered_df = filtered_df[filtered_df["GPU Max Power (W)"].isin(power_options)]
+    if server_options:
+        filtered_df = filtered_df[filtered_df["Server Name"].isin(server_options)]
 
-# Show server specifications
-if st.checkbox("Show server specifications"):
-    selected_index = st.number_input(
-        "Enter the row number to view server specifications:",
-        min_value=0,
-        max_value=len(filtered_df) - 1,
-        value=0,
-        step=1
-    )
-    selected_server = filtered_df.iloc[selected_index]["Server Name"]
-    server_specs_row = df_server_filtered[df_server_filtered["Server Name"] == selected_server]
+    # Display filtered table
+    st.dataframe(filtered_df.sort_values(by="Tokens/sec", ascending=False).reset_index(drop=True))
 
-    if not server_specs_row.empty:
-        st.subheader(f"Server specifications: {selected_server}")
-        st.table(server_specs_row.drop(columns=["Server Name"]).transpose())
-    else:
-        st.write("Specifications for the selected server are not available.")
+    # Notes
+    st.markdown("## Notes")
+    st.markdown("Click links in the 'Test Link' column for more details.")
+
+    # Server specs
+    if st.checkbox("Show server specifications"):
+        selected_index = st.number_input(
+            "Enter the row number to view server specifications:",
+            min_value=0,
+            max_value=len(filtered_df) - 1,
+            value=0,
+            step=1
+        )
+        selected_server = filtered_df.iloc[selected_index]["Server Name"]
+        server_specs_row = df_server_filtered[df_server_filtered["Server Name"] == selected_server]
+
+        if not server_specs_row.empty:
+            st.subheader(f"Server specifications: {selected_server}")
+            st.table(server_specs_row.drop(columns=["Server Name"]).transpose())
+        else:
+            st.write("Specifications for the selected server are not available.")