More refactoring

data.py

@@ -82,6 +82,33 @@ def get_sample_data() -> pd.DataFrame:
     df = df.set_index("model_name")
     return df
 
+def extract_model_data(row: pd.Series) -> tuple[dict[str, int], dict[str, int], int, int, int, int]:
+    """Extract and process model data from DataFrame row."""
+    # Handle missing values and get counts directly from dataframe
+    success_amd = int(row.get('success_amd', 0)) if pd.notna(row.get('success_amd', 0)) else 0
+    success_nvidia = int(row.get('success_nvidia', 0)) if pd.notna(row.get('success_nvidia', 0)) else 0
+    failed_multi_amd = int(row.get('failed_multi_no_amd', 0)) if pd.notna(row.get('failed_multi_no_amd', 0)) else 0
+    failed_multi_nvidia = int(row.get('failed_multi_no_nvidia', 0)) if pd.notna(row.get('failed_multi_no_nvidia', 0)) else 0
+    failed_single_amd = int(row.get('failed_single_no_amd', 0)) if pd.notna(row.get('failed_single_no_amd', 0)) else 0
+    failed_single_nvidia = int(row.get('failed_single_no_nvidia', 0)) if pd.notna(row.get('failed_single_no_nvidia', 0)) else 0
+    # Calculate total failures
+    total_failed_amd = failed_multi_amd + failed_single_amd
+    total_failed_nvidia = failed_multi_nvidia + failed_single_nvidia
+    # Create stats dictionaries directly from dataframe values
+    amd_stats = {
+        'passed': success_amd,
+        'failed': total_failed_amd,
+        'skipped': 0,  # Not available in this dataset
+        'error': 0     # Not available in this dataset
+    }
+    nvidia_stats = {
+        'passed': success_nvidia,
+        'failed': total_failed_nvidia,
+        'skipped': 0,  # Not available in this dataset
+        'error': 0     # Not available in this dataset
+    }
+    return amd_stats, nvidia_stats, failed_multi_amd, failed_single_amd, failed_multi_nvidia, failed_single_nvidia
+
 
 
 class CIResults:

model_page.py

@@ -1,6 +1,63 @@
 import matplotlib.pyplot as plt
 import pandas as pd
 from utils import generate_underlined_line
+from data import extract_model_data
+
+# Figure dimensions
+FIGURE_WIDTH_DUAL = 18
+FIGURE_HEIGHT_DUAL = 9
+
+# Colors
+COLORS = {
+    'passed': '#4CAF50',    # Medium green
+    'failed': '#E53E3E',    # More red
+    'skipped': '#FFD54F',   # Medium yellow
+    'error': '#8B0000'      # Dark red
+}
+
+# Styling constants
+BLACK = '#000000'
+LABEL_COLOR = '#AAAAAA'
+TITLE_COLOR = '#FFFFFF'
+
+# Font sizes
+DEVICE_TITLE_FONT_SIZE = 28
+
+# Layout constants
+SEPARATOR_LINE_Y_END = 0.85
+SUBPLOT_TOP = 0.85
+SUBPLOT_WSPACE = 0.4
+PIE_START_ANGLE = 90
+BORDER_LINE_WIDTH = 0.5
+SEPARATOR_ALPHA = 0.5
+SEPARATOR_LINE_WIDTH = 1
+DEVICE_TITLE_PAD = 2
+MODEL_TITLE_Y = 1
+
+# Processing constants
+MAX_FAILURE_ITEMS = 10
+
+
+def _process_failure_category(failures_obj: dict, category: str, info_lines: list) -> None:
+    """Process a single failure category (multi or single) and add to info_lines."""
+    if category in failures_obj and failures_obj[category]:
+        info_lines.append(generate_underlined_line(f"{category.title()} GPU failure details:"))
+        if isinstance(failures_obj[category], list):
+            # Handle list of failures (could be strings or dicts)
+            for i, failure in enumerate(failures_obj[category][:MAX_FAILURE_ITEMS]):
+                if isinstance(failure, dict):
+                    # Extract meaningful info from dict (e.g., test name, line, etc.)
+                    failure_str = failure.get('line', failure.get('test', 
+                                            failure.get('name', str(failure))))
+                    info_lines.append(f"  {i+1}. {failure_str}")
+                else:
+                    info_lines.append(f"  {i+1}. {str(failure)}")
+            if len(failures_obj[category]) > MAX_FAILURE_ITEMS:
+                remaining = len(failures_obj[category]) - MAX_FAILURE_ITEMS
+                info_lines.append(f"... and {remaining} more")
+        else:
+            info_lines.append(str(failures_obj[category]))
+        info_lines.append("")
 
 
 def extract_failure_info(failures_obj, device: str, multi_count: int, single_count: int) -> str:
@@ -22,193 +79,112 @@ def extract_failure_info(failures_obj, device: str, multi_count: int, single_cou
     # Try to extract detailed failure information
     try:
         if isinstance(failures_obj, dict):
-            # Check for multi and single failure categories
-            if 'multi' in failures_obj and failures_obj['multi']:
-                info_lines.append(generate_underlined_line(f"Multi GPU failure details:"))
-                if isinstance(failures_obj['multi'], list):
-                    # Handle list of failures (could be strings or dicts)
-                    for i, failure in enumerate(failures_obj['multi'][:10]):  # Limit to first 10
-                        if isinstance(failure, dict):
-                            # Extract meaningful info from dict (e.g., test name, line, etc.)
-                            failure_str = failure.get('line', failure.get('test', failure.get('name', str(failure))))
-                            info_lines.append(f"  {i+1}. {failure_str}")
-                        else:
-                            info_lines.append(f"  {i+1}. {str(failure)}")
-                    if len(failures_obj['multi']) > 10:
-                        info_lines.append(f"... and {len(failures_obj['multi']) - 10} more")
-                else:
-                    info_lines.append(str(failures_obj['multi']))
-                info_lines.append("")
-            
-            if 'single' in failures_obj and failures_obj['single']:
-                info_lines.append(generate_underlined_line(f"Single GPU failure details:"))
-                if isinstance(failures_obj['single'], list):
-                    # Handle list of failures (could be strings or dicts)
-                    for i, failure in enumerate(failures_obj['single'][:10]):  # Limit to first 10
-                        if isinstance(failure, dict):
-                            # Extract meaningful info from dict (e.g., test name, line, etc.)
-                            failure_str = failure.get('line', failure.get('test', failure.get('name', str(failure))))
-                            info_lines.append(f"  {i+1}. {failure_str}")
-                        else:
-                            info_lines.append(f"  {i+1}. {str(failure)}")
-                    if len(failures_obj['single']) > 10:
-                        info_lines.append(f"... and {len(failures_obj['single']) - 10} more")
-                else:
-                    info_lines.append(str(failures_obj['single']))
+            _process_failure_category(failures_obj, 'multi', info_lines)
+            _process_failure_category(failures_obj, 'single', info_lines)
         
         return "\n".join(info_lines) if info_lines else f"No detailed failure info for {device}"
     
     except Exception as e:
         if multi_count > 0 or single_count > 0:
-            return f"Failures detected on {device} (Multi: {multi_count}, Single: {single_count})\nDetails unavailable: {str(e)}"
+            error_msg = (f"Failures detected on {device} (Multi: {multi_count}, Single: {single_count})\n"
+                        f"Details unavailable: {str(e)}")
+            return error_msg
         return f"Error processing failure info for {device}: {str(e)}"
 
 
-def plot_model_stats(
-    df: pd.DataFrame,
-    model_name: str,
-) -> tuple[plt.Figure, str, str]:
-    """Draws a pie chart of model's passed, failed, skipped, and error stats."""
-    if df.empty or model_name not in df.index:
-        # Handle case where model data is not available
-        fig, ax = plt.subplots(figsize=(10, 8), facecolor='#000000')
-        ax.set_facecolor('#000000')
-        ax.text(0.5, 0.5, f'No data available for {model_name}', 
-                horizontalalignment='center', verticalalignment='center',
-                transform=ax.transAxes, fontsize=16, color='#888888',
-                fontfamily='monospace', weight='normal')
-        ax.set_xlim(0, 1)
-        ax.set_ylim(0, 1)
+def _create_pie_chart(ax: plt.Axes, device_label: str, filtered_stats: dict) -> None:
+    """Create a pie chart for device statistics."""
+    if not filtered_stats:
+        ax.text(0.5, 0.5, 'No test results', 
+               horizontalalignment='center', verticalalignment='center',
+               transform=ax.transAxes, fontsize=14, color='#888888',
+               fontfamily='monospace', weight='normal')
+        ax.set_title(device_label, fontsize=DEVICE_TITLE_FONT_SIZE, weight='bold', 
+                    pad=DEVICE_TITLE_PAD, color=TITLE_COLOR, fontfamily='monospace')
         ax.axis('off')
-        return fig, "No data available", "No data available"
-    
-    row = df.loc[model_name]
-    
-    # Handle missing values and get counts directly from dataframe
-    success_amd = int(row.get('success_amd', 0)) if pd.notna(row.get('success_amd', 0)) else 0
-    success_nvidia = int(row.get('success_nvidia', 0)) if pd.notna(row.get('success_nvidia', 0)) else 0
-    failed_multi_amd = int(row.get('failed_multi_no_amd', 0)) if pd.notna(row.get('failed_multi_no_amd', 0)) else 0
-    failed_multi_nvidia = int(row.get('failed_multi_no_nvidia', 0)) if pd.notna(row.get('failed_multi_no_nvidia', 0)) else 0
-    failed_single_amd = int(row.get('failed_single_no_amd', 0)) if pd.notna(row.get('failed_single_no_amd', 0)) else 0
-    failed_single_nvidia = int(row.get('failed_single_no_nvidia', 0)) if pd.notna(row.get('failed_single_no_nvidia', 0)) else 0
-    
-    # Calculate total failures
-    total_failed_amd = failed_multi_amd + failed_single_amd
-    total_failed_nvidia = failed_multi_nvidia + failed_single_nvidia
-    
-    # Softer color palette - less pastel, more vibrant
-    colors = {
-        'passed': '#4CAF50',    # Medium green
-        'failed': '#E53E3E',    # More red
-        'skipped': '#FFD54F',   # Medium yellow
-        'error': '#8B0000'      # Dark red
-    }
-    
-    # Create stats dictionaries directly from dataframe values
-    amd_stats = {
-        'passed': success_amd,
-        'failed': total_failed_amd,
-        'skipped': 0,  # Not available in this dataset
-        'error': 0     # Not available in this dataset
-    }
-    
-    nvidia_stats = {
-        'passed': success_nvidia,
-        'failed': total_failed_nvidia,
-        'skipped': 0,  # Not available in this dataset
-        'error': 0     # Not available in this dataset
-    }
-    
-    # Filter out categories with 0 values for cleaner visualization
-    amd_filtered = {k: v for k, v in amd_stats.items() if v > 0}
-    nvidia_filtered = {k: v for k, v in nvidia_stats.items() if v > 0}
-    
-    if not amd_filtered and not nvidia_filtered:
-        # Handle case where all values are 0 - minimal empty state
-        fig, ax = plt.subplots(figsize=(10, 8), facecolor='#000000')
-        ax.set_facecolor('#000000')
-        ax.text(0.5, 0.5, 'No test results available', 
-                horizontalalignment='center', verticalalignment='center',
-                transform=ax.transAxes, fontsize=16, color='#888888',
-                fontfamily='monospace', weight='normal')
-        ax.set_xlim(0, 1)
-        ax.set_ylim(0, 1)
-        ax.axis('off')
-        return fig, "", ""
-    
-    # Create figure with two subplots side by side with padding
-    fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(18, 9), facecolor='#000000')
-    ax1.set_facecolor('#000000')
-    ax2.set_facecolor('#000000')
-    
-    def create_pie_chart(ax, device_label, filtered_stats):
-        if not filtered_stats:
-            ax.text(0.5, 0.5, 'No test results', 
-                   horizontalalignment='center', verticalalignment='center',
-                   transform=ax.transAxes, fontsize=14, color='#888888',
-                   fontfamily='monospace', weight='normal')
-            ax.set_title(device_label, 
-                        fontsize=28, weight='bold', pad=2, color='#FFFFFF', 
-                        fontfamily='monospace')
-            ax.axis('off')
-            return
-            
-        chart_colors = [colors[category] for category in filtered_stats.keys()]
+        return
         
-        # Create minimal pie chart - full pie, no donut effect
-        wedges, texts, autotexts = ax.pie(
-            filtered_stats.values(), 
-            labels=[label.lower() for label in filtered_stats.keys()],  # Lowercase for minimal look
-            colors=chart_colors,
-            autopct=lambda pct: f'{int(pct/100*sum(filtered_stats.values()))}',
-            startangle=90,
-            explode=None,  # No separation
-            shadow=False,
-            wedgeprops=dict(edgecolor='#1a1a1a', linewidth=0.5),  # Minimal borders
-            textprops={'fontsize': 12, 'weight': 'normal', 'color': '#CCCCCC', 'fontfamily': 'monospace'}
-        )
+    chart_colors = [COLORS[category] for category in filtered_stats.keys()]
+    
+    # Create minimal pie chart - full pie, no donut effect
+    wedges, texts, autotexts = ax.pie(
+        filtered_stats.values(), 
+        labels=[label.lower() for label in filtered_stats.keys()],  # Lowercase for minimal look
+        colors=chart_colors,
+        autopct=lambda pct: f'{int(pct/100*sum(filtered_stats.values()))}',
+        startangle=PIE_START_ANGLE,
+        explode=None,  # No separation
+        shadow=False,
+        wedgeprops=dict(edgecolor='#1a1a1a', linewidth=BORDER_LINE_WIDTH),  # Minimal borders
+        textprops={'fontsize': 12, 'weight': 'normal', 
+                  'color': LABEL_COLOR, 'fontfamily': 'monospace'}
+    )
+    
+    # Enhanced percentage text styling for better readability
+    for autotext in autotexts:
+        autotext.set_color(BLACK)  # Black text for better contrast
+        autotext.set_weight('bold')
+        autotext.set_fontsize(14)
+        autotext.set_fontfamily('monospace')
+    
+    # Minimal category labels
+    for text in texts:
+        text.set_color(LABEL_COLOR)
+        text.set_weight('normal')
+        text.set_fontsize(13)
+        text.set_fontfamily('monospace')
+    
+    # Device label closer to chart and bigger
+    ax.set_title(device_label, fontsize=DEVICE_TITLE_FONT_SIZE, weight='normal', 
+                pad=DEVICE_TITLE_PAD, color=TITLE_COLOR, fontfamily='monospace')
+
+
+def plot_model_stats(df: pd.DataFrame, model_name: str) -> tuple[plt.Figure, str, str]:
+    """Draws pie charts of model's passed, failed, skipped, and error stats for AMD and NVIDIA."""
+    # Handle case where the dataframe is empty or the model name could not be found in it
+    if df.empty or model_name not in df.index:
+        # Create empty stats for both devices
+        amd_filtered = {}
+        nvidia_filtered = {}
+        failed_multi_amd = failed_single_amd = failed_multi_nvidia = failed_single_nvidia = 0
+        failures_amd = failures_nvidia = {}
+    else:
+        row = df.loc[model_name]
         
-        # Enhanced percentage text styling for better readability
-        for autotext in autotexts:
-            autotext.set_color('#000000')  # Black text for better contrast
-            autotext.set_weight('bold')
-            autotext.set_fontsize(14)
-            autotext.set_fontfamily('monospace')
+        # Extract and process model data
+        amd_stats, nvidia_stats, failed_multi_amd, failed_single_amd, failed_multi_nvidia, failed_single_nvidia = \
+            extract_model_data(row)
         
-        # Minimal category labels
-        for text in texts:
-            text.set_color('#AAAAAA')
-            text.set_weight('normal')
-            text.set_fontsize(13)
-            text.set_fontfamily('monospace')
+        # Filter out categories with 0 values for cleaner visualization
+        amd_filtered = {k: v for k, v in amd_stats.items() if v > 0}
+        nvidia_filtered = {k: v for k, v in nvidia_stats.items() if v > 0}
         
-        # Device label closer to chart and bigger
-        ax.set_title(device_label, 
-                    fontsize=28, weight='normal', pad=2, color='#FFFFFF', 
-                    fontfamily='monospace')
+        # Generate failure info directly from dataframe
+        failures_amd = row.get('failures_amd', {})
+        failures_nvidia = row.get('failures_nvidia', {})
+    
+    # Always create figure with two subplots side by side with padding
+    fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(FIGURE_WIDTH_DUAL, FIGURE_HEIGHT_DUAL), facecolor=BLACK)
+    ax1.set_facecolor(BLACK)
+    ax2.set_facecolor(BLACK)
     
     # Create both pie charts with device labels
-    create_pie_chart(ax1, "amd", amd_filtered)
-    create_pie_chart(ax2, "nvidia", nvidia_filtered)
+    _create_pie_chart(ax1, "amd", amd_filtered)
+    _create_pie_chart(ax2, "nvidia", nvidia_filtered)
     
     # Add subtle separation line between charts - stops at device labels level
     line_x = 0.5
-    fig.add_artist(plt.Line2D([line_x, line_x], [0.0, 0.85], 
-                              color='#333333', linewidth=1, alpha=0.5,
-                              transform=fig.transFigure))
+    fig.add_artist(plt.Line2D([line_x, line_x], [0.0, SEPARATOR_LINE_Y_END], 
+                              color='#333333', linewidth=SEPARATOR_LINE_WIDTH, 
+                              alpha=SEPARATOR_ALPHA, transform=fig.transFigure))
     
     # Add central shared title for model name
-    fig.suptitle(f'{model_name.lower()}', 
-                fontsize=32, weight='bold', color='#CCCCCC', 
-                fontfamily='monospace', y=1)
+    fig.suptitle(f'{model_name.lower()}', fontsize=32, weight='bold', 
+                color='#CCCCCC', fontfamily='monospace', y=MODEL_TITLE_Y)
     
     # Clean layout with padding and space for central title
     plt.tight_layout()
-    plt.subplots_adjust(top=0.85, wspace=0.4)  # Added wspace for padding between charts
-    
-    # Generate failure info directly from dataframe
-    failures_amd = row.get('failures_amd', {})
-    failures_nvidia = row.get('failures_nvidia', {})
+    plt.subplots_adjust(top=SUBPLOT_TOP, wspace=SUBPLOT_WSPACE)
     
     amd_failed_info = extract_failure_info(failures_amd, 'AMD', failed_multi_amd, failed_single_amd)
     nvidia_failed_info = extract_failure_info(failures_nvidia, 'NVIDIA', failed_multi_nvidia, failed_single_nvidia)

summary_page.py

@@ -1,5 +1,6 @@
 import matplotlib.pyplot as plt
 import pandas as pd
+from data import extract_model_data
 
 # Layout parameters
 COLUMNS = 3
@@ -95,28 +96,8 @@ def create_summary_page(df: pd.DataFrame, available_models: list[str]) -> plt.Fi
             
         row = df.loc[model_name]
         
-        # Get values directly from dataframe
-        success_amd = int(row.get('success_amd', 0)) if pd.notna(row.get('success_amd', 0)) else 0
-        success_nvidia = int(row.get('success_nvidia', 0)) if pd.notna(row.get('success_nvidia', 0)) else 0
-        failed_multi_amd = int(row.get('failed_multi_no_amd', 0)) if pd.notna(row.get('failed_multi_no_amd', 0)) else 0
-        failed_multi_nvidia = int(row.get('failed_multi_no_nvidia', 0)) if pd.notna(row.get('failed_multi_no_nvidia', 0)) else 0
-        failed_single_amd = int(row.get('failed_single_no_amd', 0)) if pd.notna(row.get('failed_single_no_amd', 0)) else 0
-        failed_single_nvidia = int(row.get('failed_single_no_nvidia', 0)) if pd.notna(row.get('failed_single_no_nvidia', 0)) else 0
-        
-        # Calculate stats
-        amd_stats = {
-            'passed': success_amd,
-            'failed': failed_multi_amd + failed_single_amd,
-            'skipped': 0,
-            'error': 0
-        }
-        
-        nvidia_stats = {
-            'passed': success_nvidia,
-            'failed': failed_multi_nvidia + failed_single_nvidia,
-            'skipped': 0,
-            'error': 0
-        }
+        # Extract and process model data
+        amd_stats, nvidia_stats = extract_model_data(row)[:2]
             
         # Calculate position in 4-column grid
         col = visible_model_count % COLUMNS