plot

name	plot
description	Plot a bubble chart from an already-aggregated results CSV. Use when the user wants to visualize model performance as a bubble chart and already has a CSV file (e.g. produced by the `aggregate` skill or `tool/aggregate_results.py`).

When to use / what to ask

The user has an aggregated CSV (typically work_dirs/<dataset>/results_all.csv) and wants a bubble chart. If the CSV does not exist yet, run the aggregate skill first.

Ask for the CSV path and the three axis columns: --x / --y / --size (e.g. mse / mae / total_params).

Command

uv run python tool/plot_bubble.py \
    --csv work_dirs/ETTh1/results_all.csv \
    --x mse --y mae --size total_params

Output defaults to work_dirs/plots/bubble_<csv-stem>.svg. Scales (--x-scale/--y-scale/--size-scale), grouping (--color-by/--label-by), --title, and --output are tunable — see --help.

tsf aggregate-plot runs aggregation + plotting in one step: uv run python tool/tsf.py aggregate-plot --dataset <name> --pred-len <len>.

Reference

Full flags and behavior (log scales silently drop non-positive rows; bubble sizes normalized to 30–300 pt²): docs/en/plot-bubble.md.