Plot Gallery for Multipath

This page collects all published metric plots for the experiment 2026-06-18_22-42-17_403726-multipath on node Hactar. The plots are generated from the enrichment manifests and published here as self-contained browser-friendly PNG galleries for reviewer inspection.

Back to node landing page Visualization website of the enrichment runs

Overview and Combined Summary

Combined Output Tokens and Elapsed Time — Combined Output Tokens and Elapsed Time

Compares output length and total runtime across all included experiments and tested models.

combined_output_tokens_elapsed_seconds.png

Quality

Normalized BERTScore F1 — Normalized BERTScore F1

Shows how the mean normalized BERTScore F1 values vary across the published runs.

boxplot_bertscore_f1_mean.png

Entity BERTScore F1 Distribution — Entity BERTScore F1 Distribution

Shows how the entity-aware BERTScore F1 values vary across the published runs.

boxplot_entity_bertscore_f1.png

Runtime and Throughput

Elapsed Time — Elapsed Time

Shows the total runtime distribution across the published runs.

boxplot_elapsed_seconds.png

Generation Time — Generation Time

Shows how long text generation takes across the published runs.

boxplot_eval_seconds.png

Model Load Time — Model Load Time

Shows the model loading time distribution across the published runs.

boxplot_load_seconds.png

Output Tokens per Second — Output Tokens per Second

Shows the text-generation throughput across the published runs.

boxplot_output_tokens_per_second.png

Prompt Evaluation Time — Prompt Evaluation Time

Shows how long prompt evaluation takes across the published runs.

boxplot_prompt_eval_seconds.png

Prompt Tokens per Second — Prompt Tokens per Second

Shows the prompt-processing throughput across the published runs.

boxplot_prompt_tokens_per_second.png

Token Usage

Output Tokens — Output Tokens

Shows the output token counts across the published runs.

boxplot_output_tokens.png

Prompt Tokens — Prompt Tokens

Shows the prompt token counts across the published runs.

boxplot_prompt_tokens.png

GPU, Energy, and Temperature

GPU Utilization — GPU Utilization

Shows the average GPU utilization observed during the published runs.

boxplot_gpu_utilization_avg_pct.png

Average Power Draw — Average Power Draw

Shows the average GPU power draw observed during the published runs.

boxplot_power_draw_avg_watts.png

Peak Power Draw — Peak Power Draw

Shows the peak GPU power draw observed during the published runs.

boxplot_power_draw_max_watts.png

Peak Temperature — Peak Temperature

Shows the peak GPU temperature observed during the published runs.

boxplot_temperature_max_c.png