Visualization

Standalone plot functions (module-level). Trainer convenience wrappers such as trainer.plot_training_convergence() are documented on Trainer and TextPredictionTrainer.

Training and encoder plots

• plot_training_convergence — Convergence over training steps
• plot_encoder_distributions — Split violins of encoded values
• plot_encoder_scatter — Interactive encoder scatter (Plotly)
• plot_encoder_by_target — Encoded values per masked target token

Multi-model comparison

• plot_topk_overlap_heatmap — Pairwise top-k weight overlap
• plot_topk_overlap_venn — Top-k overlap Venn diagram
• plot_similarity_heatmap — Similarity matrix heatmap
• plot_cross_encoding_heatmap — Oriented cross-encoding heatmap
• plot_gradiend_transition_cross_encoding_heatmap — GRADIEND × transition heatmap
• plot_gradiend_feature_cross_encoding_heatmap — GRADIEND × feature-class heatmap
• plot_comparison_heatmap — Generic comparison heatmap (e.g. seed comparison)

Environment

• check_plot_environment — Verify matplotlib/LaTeX/font setup
• configure_plot_style — Apply GRADIEND matplotlib defaults
• PlotStyleConfig — LaTeX, font, preamble, and transition-arrow options
• format_transition_label — Render A -> B labels for heatmaps

See also Plot styling & LaTeX guide.

Related

• compute_similarity_matrix — Build similarity data for heatmaps
• Evaluation visualization guide — Plot customization and examples