Update app.py

app.py

@@ -2,107 +2,157 @@ import streamlit as st
 import pandas as pd
 from PIL import Image, ImageDraw, ImageFont
 import io
-import os, socket
+import os
+import socket
+import calendar
+import re
+from typing import Optional
+from huggingface_hub import hf_hub_download
+
+# =========================
+# Hugging Face Space config
+# =========================
+HF_REPO_ID = "AIEnergyScore/Leaderboard"   # Space slug
+HF_REPO_TYPE = "space"
+HF_DATA_PREFIX = "data/energy"             # path within the Space
 
-# --- CSV mapping (HF Space: AIEnergyScore/Leaderboard -> data/energy) ---
+# =========================
+# Task -> CSV mapping
+# =========================
 TASK_TO_CSV = {
-    "Image Generation":     "image_generation.csv",
-    "Text Classification":  "text_classification.csv",
-    "Image Classification": "image_classification.csv",
-    "Image Captioning":     "image_captioning.csv",
-    "Summarization":        "summarization.csv",
-    "Text Generation":      "text_generation.csv",
-    "Reasoning":            "reasoning.csv",
+    "Text Generation":        "text_generation.csv",
+    "Reasoning":              "reasoning.csv",           # now exists in your Space
+    "Image Generation":       "image_generation.csv",
+    "Text Classification":    "text_classification.csv",
+    "Image Classification":   "image_classification.csv",
+    "Image Captioning":       "image_captioning.csv",
+    "Summarization":          "summarization.csv",
+    "Speech-to-Text (ASR)":   "asr.csv",
+    "Object Detection":       "object_detection.csv",
+    "Question Answering":     "question_answering.csv",
+    "Sentence Similarity":    "sentence_similarity.csv",
 }
-
-# Back-compat so old code like task_to_file[task] keeps working
+# Back-compat if parts of the code still reference this name:
 task_to_file = TASK_TO_CSV
 
-from huggingface_hub import hf_hub_download
-
-HF_REPO_ID = "AIEnergyScore/Leaderboard"   # this is a Space
-HF_REPO_TYPE = "space"
-HF_DATA_PREFIX = "data/energy"             # where the CSVs live in the Space
-
+# =========================
+# Helpers
+# =========================
 def read_csv_from_hub(file_name: str) -> pd.DataFrame:
     """
-    Download a CSV from the HF Space path data/energy/<file_name>
-    and return it as a pandas DataFrame.
-    Falls back to local file if Hub is unavailable.
+    Download a CSV from HF Space path data/energy/<file_name>,
+    return a pandas DataFrame. Falls back to local if hub unavailable.
     """
     hub_path = f"{HF_DATA_PREFIX}/{file_name}"
     try:
-        # fast DNS sanity check to give a useful error if DNS is down
+        # helpful DNS check
         socket.gethostbyname("huggingface.co")
-
         local_path = hf_hub_download(
             repo_id=HF_REPO_ID,
             repo_type=HF_REPO_TYPE,
             filename=hub_path,
-            revision="main",          # adjust if you use a branch/tag
+            revision="main",
             resume_download=True
         )
         return pd.read_csv(local_path)
     except Exception as e:
-        # Fallback to local file if present
         try:
             return pd.read_csv(file_name)
         except Exception:
-            # re-raise with context so it surfaces nicely in Streamlit
             raise RuntimeError(
                 f"Unable to load '{file_name}' from Hub path '{hub_path}' or locally. "
                 f"Original error: {e}"
             )
 
+def _normalize(col: str) -> str:
+    return re.sub(r"[^a-z0-9]", "", col.strip().lower())
+
+def find_test_date_column(df: pd.DataFrame) -> Optional[str]:
+    """
+    Locate a 'test date' column. Strategy:
+    1) Exact case-insensitive match 'test date'
+    2) Any header whose normalized form contains both 'test' and 'date'
+    3) Fallback to column E (index 4) if present
+    """
+    # (1) exact "test date"
+    for c in df.columns:
+        if c.strip().lower() == "test date":
+            return c
+    # (2) flexible match
+    for c in df.columns:
+        cn = _normalize(c)
+        if "test" in cn and "date" in cn:
+            return c
+    # (3) fallback to E (0-based index 4)
+    if len(df.columns) >= 5:
+        return df.columns[4]
+    return None
+
+def month_abbrev_to_full(abbrev: str) -> Optional[str]:
+    """
+    Map 'Feb' -> 'February', 'Oct' -> 'October'. Returns None if unknown.
+    """
+    if not isinstance(abbrev, str) or not abbrev:
+        return None
+    abbr = abbrev.strip()[:3].title()  # normalize to 3-letter case 'Oct'
+    for m in range(1, 13):
+        if calendar.month_abbr[m] == abbr:
+            return calendar.month_name[m]
+    return None
+
+def render_date_from_test_date(value: str) -> str:
+    """
+    Accepts 'Oct 2025', 'Feb 2025' and returns 'October 2025', 'February 2025'.
+    Returns '' if it can’t parse.
+    """
+    if not isinstance(value, str):
+        return ""
+    s = value.strip()
+    m = re.match(r"^([A-Za-z]+)\s+(\d{4})$", s)
+    if not m:
+        return ""
+    month_full = month_abbrev_to_full(m.group(1))
+    return f"{month_full} {m.group(2)}" if month_full else ""
+
+def smart_capitalize(text):
+    """Capitalize first letter only if not already; leave rest unchanged."""
+    if not text:
+        return text
+    return text if text[0].isupper() else text[0].upper() + text[1:]
+
+# =========================
+# UI / App
+# =========================
 def main():
-    # Inject custom CSS to change the color of selected tasks
+    # Tag styling
     st.markdown(
         """
         <style>
-            /* Change background color of selected items */
             .stMultiSelect [data-baseweb="tag"] {
-                background-color: #3fa45bff !important; /* Custom green */
-                color: white !important; /* White text */
-                font-weight: medium;
+                background-color: #3fa45bff !important;
+                color: white !important;
+                font-weight: 500;
                 border-radius: 5px;
                 padding: 5px 10px;
             }
-
-            /* Change hover effect */
-            .stMultiSelect [data-baseweb="tag"]:hover {
-                background-color: #358d4d !important;
-            }
-
-            /* Style the dropdown input field */
-            .stMultiSelect input {
-                color: black !important;
-            }
+            .stMultiSelect [data-baseweb="tag"]:hover { background-color: #358d4d !important; }
+            .stMultiSelect input { color: black !important; }
         </style>
         """,
         unsafe_allow_html=True,
     )
 
-    # Sidebar logo and title
+    # Sidebar logo & title
     with st.sidebar:
         col1, col2 = st.columns([1, 5])
-
         with col1:
             logo = Image.open("logo.png")
-            resized_logo = logo.resize((50, 50))
-            st.image(resized_logo)
-
+            st.image(logo.resize((50, 50)))
         with col2:
             st.markdown(
                 """
-                <div style="
-                    display: flex; 
-                    align-items: center; 
-                    gap: 10px; 
-                    margin: 0; 
-                    padding: 0; 
-                    font-family: 'Inter', sans-serif; 
-                    font-size: 26px; 
-                    font-weight: medium;">
+                <div style="display:flex;align-items:center;gap:10px;margin:0;padding:0;
+                            font-family:'Inter',sans-serif;font-size:26px;font-weight:500;">
                     AI Energy Score
                 </div>
                 """,
@@ -111,38 +161,27 @@ def main():
 
     st.sidebar.markdown("<hr style='border: 1px solid gray; margin: 15px 0;'>", unsafe_allow_html=True)
     st.sidebar.write("### Generate Label:")
-    # Define the ordered list of tasks.
+
+    # Task order
     task_order = [
-        "Text Generation", 
-        "Reasoning", 
-        "Image Generation", 
-        "Text Classification", 
-        "Image Classification", 
-        "Image Captioning", 
-        "Summarization", 
-        "Speech-to-Text (ASR)", 
-        "Object Detection", 
-        "Question Answering", 
-        "Sentence Similarity"
+        "Text Generation",
+        "Reasoning",
+        "Image Generation",
+        "Text Classification",
+        "Image Classification",
+        "Image Captioning",
+        "Summarization",
+        "Speech-to-Text (ASR)",
+        "Object Detection",
+        "Question Answering",
+        "Sentence Similarity",
     ]
-    # Task selection
+
+    # 1) Select task(s)
     st.sidebar.write("#### 1. Select task(s) to view models")
     selected_tasks = st.sidebar.multiselect("", options=task_order, default=["Text Generation"])
-    # Mapping from task to CSV file name.
-    TASK_TO_CSV = {
-        "Text Generation": "text_generation.csv",
-        "Reasoning": "reasoning.csv",
-        "Image Generation": "image_generation.csv",
-        "Text Classification": "text_classification.csv",
-        "Image Classification": "image_classification.csv",
-        "Image Captioning": "image_captioning.csv",
-        "Summarization": "summarization.csv",
-        "Speech-to-Text (ASR)": "asr.csv",
-        "Object Detection": "object_detection.csv",
-        "Question Answering": "question_answering.csv",
-        "Sentence Similarity": "sentence_similarity.csv"
-    }
-    st.sidebar.write("#### 2. Select a model to generate label")
+
+    # Default when nothing selected
     default_model_data = {
         'provider': "AI Provider",
         'model': "Model Name",
@@ -150,15 +189,20 @@ def main():
         'date': "",
         'task': "",
         'hardware': "",
-        'energy': "?",
+        'energy': 0.0,
         'score': 5
     }
+
     if not selected_tasks:
         model_data = default_model_data
     else:
         dfs = []
         for task in selected_tasks:
-            file_name = task_to_file[task]
+            file_name = TASK_TO_CSV.get(task)
+            if not file_name:
+                st.sidebar.error(f"Unknown task '{task}'.")
+                continue
+
             try:
                 df = read_csv_from_hub(file_name)
             except FileNotFoundError:
@@ -168,15 +212,31 @@ def main():
                 st.sidebar.error(f"Error reading '{file_name}' for task {task}: {e}")
                 continue
 
+            # Split provider/model if combined as "Provider/Model"
             df['full_model'] = df['model']
             df[['provider', 'model']] = df['model'].str.split(pat='/', n=1, expand=True)
-            # Multiply raw energy by 1000 to convert to Wh, then round to 2 decimals
+
+            # Convert kWh -> Wh (total_gpu_energy is in kWh); keep 2 decimals
             df['energy'] = (df['total_gpu_energy'] * 1000).round(2)
+
+            # Score
             df['score'] = df['energy_score'].fillna(1).astype(int)
-            df['date'] = "February 2025"
+
+            # Hardware placeholder (adjust if you have a specific column)
             df['hardware'] = "NVIDIA H100-80GB"
             df['task'] = task
 
+            # --- DATE: Use CSV 'test date' for Text Generation & Reasoning ---
+            if task in {"Text Generation", "Reasoning"}:
+                td_col = find_test_date_column(df)
+                if td_col:
+                    df['date'] = df[td_col].apply(render_date_from_test_date)
+                    df['date'] = df['date'].fillna("").astype(str)
+                else:
+                    df['date'] = ""
+            else:
+                df['date'] = ""
+
             dfs.append(df)
 
         if not dfs:
@@ -195,7 +255,7 @@ def main():
                 model_data = data_df[data_df["full_model"] == selected_model].iloc[0]
 
     st.sidebar.write("#### 3. Download the label")
-    
+
     try:
         score = int(model_data["score"])
         background_path = f"{score}.png"
@@ -228,21 +288,21 @@ def main():
     st.sidebar.markdown("<hr style='border: 1px solid gray; margin: 15px 0;'>", unsafe_allow_html=True)
     st.sidebar.write("### Key Links")
     st.sidebar.markdown(
-    """
-    <ul style="margin-top: 0; margin-bottom: 0; padding-left: 20px;">
-        <li><a href="https://huggingface.co/spaces/AIEnergyScore/Leaderboard" target="_blank">Leaderboard</a></li>
-        <li><a href="https://huggingface.co/spaces/AIEnergyScore/submission_portal" target="_blank">Submission Portal</a></li>
-        <li><a href="https://huggingface.github.io/AIEnergyScore/#faq" target="_blank">FAQ</a></li>
-        <li><a href="https://huggingface.github.io/AIEnergyScore/#documentation" target="_blank">Documentation</a></li>
-    </ul>
-    """,
-    unsafe_allow_html=True,
+        """
+        <ul style="margin-top:0;margin-bottom:0;padding-left:20px;">
+            <li><a href="https://huggingface.co/spaces/AIEnergyScore/Leaderboard" target="_blank">Leaderboard</a></li>
+            <li><a href="https://huggingface.co/spaces/AIEnergyScore/submission_portal" target="_blank">Submission Portal</a></li>
+            <li><a href="https://huggingface.github.io/AIEnergyScore/#faq" target="_blank">FAQ</a></li>
+            <li><a href="https://huggingface.github.io/AIEnergyScore/#documentation" target="_blank">Documentation</a></li>
+        </ul>
+        """,
+        unsafe_allow_html=True,
     )
 
 def create_label_single_pass(background_image, model_data, final_size=(520, 728)):
     bg_resized = background_image.resize(final_size, Image.Resampling.LANCZOS)
 
-    # If no task is selected (i.e. using default model_data), return the background without drawing any text.
+    # If no task is selected (i.e., using default model_data), return background
     if not model_data.get("task"):
         return bg_resized
 
@@ -258,34 +318,33 @@ def create_label_single_pass(background_image, model_data, final_size=(520, 728)
 
     title_x, title_y = 33, 150
     details_x, details_y = 480, 256
-    energy_x = 480  # Right margin for the energy value
-    energy_y = 472  
+    energy_x, energy_y = 480, 472  # right-aligned anchors
 
-    # Capitalize only the first letter of the first word while keeping the rest as is
-    def smart_capitalize(text):
-        """Capitalizes the first letter of a string only if it's not already capitalized."""
-        if not text:
-            return text  # Return unchanged if empty
-        return text if text[0].isupper() else text[0].upper() + text[1:]
-
-    # Apply smart capitalization
     provider_text = smart_capitalize(str(model_data['provider']))
     model_text = smart_capitalize(str(model_data['model']))
-    
+
     draw.text((title_x, title_y), provider_text, font=title_font, fill="black")
     draw.text((title_x, title_y + 38), model_text, font=title_font, fill="black")
 
-    details_lines = [str(model_data['date']), str(model_data['task']), str(model_data['hardware'])]
+    # Right-align details lines (date, task, hardware)
+    details_lines = [
+        str(model_data.get('date', "")),
+        str(model_data.get('task', "")),
+        str(model_data.get('hardware', "")),
+    ]
     for i, line in enumerate(details_lines):
         bbox = draw.textbbox((0, 0), line, font=details_font)
-        text_width = bbox[2] - bbox[0]  # Get text width
+        text_width = bbox[2] - bbox[0]
         draw.text((details_x - text_width, details_y + i * 47), line, font=details_font, fill="black")
 
-    # Format the energy value to 2 decimal places and right-align the text
-    energy_text = f"{model_data['energy']:.2f}"
+    # Energy value (two decimals) right-aligned
+    try:
+        energy_value = float(model_data.get('energy', 0.0))
+    except Exception:
+        energy_value = 0.0
+    energy_text = f"{energy_value:.2f}"
     energy_bbox = draw.textbbox((0, 0), energy_text, font=energy_font)
     energy_text_width = energy_bbox[2] - energy_bbox[0]
-
     draw.text((energy_x - energy_text_width, energy_y), energy_text, font=energy_font, fill="black")
 
     return bg_resized