Download components/stability_scatter.py from ACA050/aegislm: direct link, hf CLI and curl.
- Browser
- Download file 8 kB
-
https://huggingface.co/spaces/ACA050/aegislm/resolve/main/components/stability_scatter.py
- Command line
-
hf download hf://spaces/ACA050/aegislm/components/stability_scatter.py
-
curl -L -o stability_scatter.py https://huggingface.co/spaces/ACA050/aegislm/resolve/main/components/stability_scatter.py
8 kB
| """ | |
| Stability Scatter Plot Component | |
| Visualization component for stability analysis between baseline and adversarial robustness. | |
| Shows models plotted on R_base vs R_adv space with diagonal reference line. | |
| """ | |
| import logging | |
| from typing import Any, List, Optional | |
| import plotly.graph_objects as go | |
| from dashboard.schemas import BenchmarkComparisonData, BenchmarkModelResult | |
| from dashboard.utils import log_dashboard_event | |
| logger = logging.getLogger(__name__) | |
| def create_stability_scatter( | |
| comparison: Optional[BenchmarkComparisonData], | |
| ) -> Any: | |
| """ | |
| Create stability scatter plot from benchmark comparison data. | |
| Args: | |
| comparison: BenchmarkComparisonData object | |
| Returns: | |
| Plotly figure | |
| """ | |
| if comparison is None or not comparison.model_results: | |
| return create_empty_stability_chart("No benchmark data available") | |
| log_dashboard_event( | |
| "DASHBOARD_VIEW_STABILITY_CHART", | |
| benchmark_id=comparison.benchmark_id, | |
| ) | |
| return update_stability_scatter(comparison) | |
| def update_stability_scatter( | |
| comparison: BenchmarkComparisonData, | |
| ) -> Any: | |
| """ | |
| Update stability scatter plot with benchmark comparison data. | |
| Each model is plotted as a point at (R_base, R_adv). | |
| A diagonal line y=x shows perfect stability. | |
| Closer to diagonal = more stable. | |
| Args: | |
| comparison: BenchmarkComparisonData object | |
| Returns: | |
| Plotly figure | |
| """ | |
| if not comparison.model_results: | |
| return create_empty_stability_chart("No model results") | |
| # Extract data | |
| baselines = [r.baseline_robustness for r in comparison.model_results] | |
| adversarials = [r.adversarial_robustness for r in comparison.model_results] | |
| models = [r.model_name for r in comparison.model_results] | |
| rsis = [r.rsi for r in comparison.model_results] | |
| # Color based on RSI (stability) | |
| # RSI close to 1 = stable = green | |
| # RSI far from 1 = unstable = red | |
| colors = [] | |
| for rsi in rsis: | |
| # RSI range: 0 to 2, target is 1 | |
| distance_from_1 = abs(rsi - 1) | |
| if distance_from_1 < 0.1: | |
| colors.append("#22c55e") # Green - very stable | |
| elif distance_from_1 < 0.3: | |
| colors.append("#84cc16") # Lime - stable | |
| elif distance_from_1 < 0.5: | |
| colors.append("#eab308") # Yellow - moderate | |
| elif distance_from_1 < 0.7: | |
| colors.append("#f97316") # Orange - unstable | |
| else: | |
| colors.append("#ef4444") # Red - very unstable | |
| # Create scatter plot | |
| fig = go.Figure() | |
| # Add diagonal line (perfect stability) | |
| fig.add_trace( | |
| go.Scatter( | |
| x=[0, 1], | |
| y=[0, 1], | |
| mode="lines", | |
| name="Perfect Stability (y=x)", | |
| line=dict(color="gray", dash="dash", width=2), | |
| hoverinfo="name", | |
| ) | |
| ) | |
| # Add model points | |
| fig.add_trace( | |
| go.Scatter( | |
| x=baselines, | |
| y=adversarials, | |
| mode="markers+text", | |
| name="Models", | |
| marker=dict( | |
| size=14, | |
| color=colors, | |
| line=dict(color="white", width=1), | |
| ), | |
| text=models, | |
| textposition="top center", | |
| textfont=dict(size=10), | |
| hovertemplate=( | |
| "<b>%{text}</b><br>" | |
| "R_base: %{x:.4f}<br>" | |
| "R_adv: %{y:.4f}<br>" | |
| "<extra></extra>" | |
| ), | |
| ) | |
| ) | |
| # Layout | |
| fig.update_layout( | |
| title={ | |
| "text": f"Robustness Stability - {comparison.benchmark_name}", | |
| "x": 0.5, | |
| "xanchor": "center", | |
| }, | |
| xaxis=dict( | |
| title="R_base (Baseline Robustness)", | |
| range=[0, 1], | |
| tickformat=".2f", | |
| constrain="domain", | |
| ), | |
| yaxis=dict( | |
| title="R_adv (Adversarial Robustness)", | |
| range=[0, 1], | |
| tickformat=".2f", | |
| scaleanchor="x", | |
| scaleratio=1, | |
| ), | |
| height=500, | |
| width=600, | |
| legend=dict( | |
| yanchor="top", | |
| y=0.99, | |
| xanchor="left", | |
| x=0.01, | |
| ), | |
| hovermode="closest", | |
| ) | |
| return fig | |
| def create_empty_stability_chart(message: str = "Select a benchmark") -> Any: | |
| """ | |
| Create empty stability chart with message. | |
| Args: | |
| message: Message to display | |
| Returns: | |
| Plotly figure | |
| """ | |
| fig = go.Figure() | |
| # Add diagonal line | |
| fig.add_trace( | |
| go.Scatter( | |
| x=[0, 1], | |
| y=[0, 1], | |
| mode="lines", | |
| name="Perfect Stability (y=x)", | |
| line=dict(color="gray", dash="dash", width=2), | |
| ) | |
| ) | |
| fig.update_layout( | |
| title={ | |
| "text": f"Robustness Stability - {message}", | |
| "x": 0.5, | |
| "xanchor": "center", | |
| }, | |
| xaxis=dict( | |
| title="R_base", | |
| range=[0, 1], | |
| ), | |
| yaxis=dict( | |
| title="R_adv", | |
| range=[0, 1], | |
| scaleanchor="x", | |
| scaleratio=1, | |
| ), | |
| height=500, | |
| width=600, | |
| ) | |
| return fig | |
| def create_stability_scatter_from_results( | |
| results: List[BenchmarkModelResult], | |
| title: str = "Robustness Stability", | |
| ) -> Any: | |
| """ | |
| Create stability scatter plot from list of model results. | |
| Args: | |
| results: List of BenchmarkModelResult objects | |
| title: Chart title | |
| Returns: | |
| Plotly figure | |
| """ | |
| if not results: | |
| return create_empty_stability_chart("No results") | |
| # Extract data | |
| baselines = [r.baseline_robustness for r in results] | |
| adversarials = [r.adversarial_robustness for r in results] | |
| models = [r.model_name for r in results] | |
| rsis = [r.rsi for r in results] | |
| # Color based on RSI | |
| colors = [] | |
| for rsi in rsis: | |
| distance_from_1 = abs(rsi - 1) | |
| if distance_from_1 < 0.1: | |
| colors.append("#22c55e") | |
| elif distance_from_1 < 0.3: | |
| colors.append("#84cc16") | |
| elif distance_from_1 < 0.5: | |
| colors.append("#eab308") | |
| elif distance_from_1 < 0.7: | |
| colors.append("#f97316") | |
| else: | |
| colors.append("#ef4444") | |
| fig = go.Figure() | |
| # Add diagonal line | |
| fig.add_trace( | |
| go.Scatter( | |
| x=[0, 1], | |
| y=[0, 1], | |
| mode="lines", | |
| name="Perfect Stability (y=x)", | |
| line=dict(color="gray", dash="dash", width=2), | |
| ) | |
| ) | |
| # Add model points | |
| fig.add_trace( | |
| go.Scatter( | |
| x=baselines, | |
| y=adversarials, | |
| mode="markers+text", | |
| name="Models", | |
| marker=dict( | |
| size=14, | |
| color=colors, | |
| line=dict(color="white", width=1), | |
| ), | |
| text=models, | |
| textposition="top center", | |
| textfont=dict(size=10), | |
| ) | |
| ) | |
| fig.update_layout( | |
| title={ | |
| "text": title, | |
| "x": 0.5, | |
| "xanchor": "center", | |
| }, | |
| xaxis=dict( | |
| title="R_base", | |
| range=[0, 1], | |
| ), | |
| yaxis=dict( | |
| title="R_adv", | |
| range=[0, 1], | |
| scaleanchor="x", | |
| scaleratio=1, | |
| ), | |
| height=500, | |
| width=600, | |
| legend=dict( | |
| yanchor="top", | |
| y=0.99, | |
| xanchor="left", | |
| x=0.01, | |
| ), | |
| ) | |
| return fig | |