Repository navigation
Explorers: saved views, chart size, hidden unselected methods, a win-rate title and renames - #688
Merged
Merged
Conversation
…rate title and renames The Pareto and tuning-trajectories explorers take a typed chart width and height, and can hide the unselected methods instead of greying them out. The win-rate explorer gets an editable figure title drawn in the matrix (opening on the subset's name) and renames models from their chips. Each of them has a "Save view" button that downloads the page opening on the current view, still editable. `generate_all_figs` also writes an overall tuning-trajectories explorer with a subset selector, next to the composite Pareto explorer. `generate_all_figs(static_figures=...)` defaults to "uncovered": the static figures an explorer shows and exports (the four `pareto_front_*` figures, the win-rate matrix figure, and the tuning-trajectory figures of Improvability or Elo against train, inference or total time) are written only with "all". Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01ELdutHiUqkvynzEP7EnPsi
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
The explorers get more figure controls: the Pareto and tuning-trajectories explorers take a typed chart width and height and can hide the unselected methods instead of greying them out, and the win-rate explorer gets an editable figure title drawn in the matrix and renames models from their chips. Every one of them can save its current view as a new, still editable page.
generate_all_figsalso writes an overall tuning-trajectories explorer with a subset selector, and leaves out by default the static figures an explorer already shows and exports.Details
Pareto and tuning-trajectories explorers:
build_pareto_explorer_html(hide_unselected=True)opens a page that way.Win-rate explorer:
build_winrate_explorer_html(method_names=...)presets names.Saved views:
<script type="application/json" id="saved-view">block; the page applies a saved view before its first render, so the file opens on that view and stays fully editable, and saving it again replaces the block.generate_all_figs:save_composite_trajectories_explorer(on wheneverplot_tuning_trajectoriesis) writes<output_dir>/tuning_trajectories_explorer.htmlfrom the subsets'tuning_trajectories/tuning_trajectories.csv, as the composite Pareto explorer is built from theirpareto_front_points.csv.static_figuresdefaults to"uncovered", which skips the static figures an explorer shows and exports: each subset's fourpareto_front_*figures, its win-rate matrix figure, and the nine tuning-trajectory figures of Improvability or Elo against train, inference or total time."all"writes them too. Figures without an interactive page (the tuning-impact bar plots, the composite leaderboard, the trajectory diagnostics such as baseline advantage, metric error and validation against test) are written either way.LeaderboardReporter.evaland the tuning-trajectory functions take the same argument with"all"as the default, so direct callers are unchanged. On a 14-subset run with trajectories the default writes 238 figures instead of 434.Tests
ruff check . && ruff format --check .(ruff 0.14.0)pytest tests/tabarena/contexts tests/tabarena/plot tests/tabarena/evaluation tests/tabarena/website -q: 399 passed, with the newtest_a_view_is_saved_into_a_copy_of_the_page_as_it_loaded,test_the_chart_size_is_typed_for_the_page,test_unselected_methods_can_be_hidden,test_the_figure_title_is_drawn_in_the_matrix_and_edited,test_models_are_renamed_for_display_only(win-rate),test_the_composite_trajectories_explorer_holds_every_subset,test_static_figures_reach_the_compare_and_trajectory_passes,test_without_static_figures_only_the_explorer_pages_are_writtenandtest_uncovered_static_figures_leave_out_the_explorer_views.generate_all_figsrun:🤖 Generated with Claude Code
https://claude.ai/code/session_01ELdutHiUqkvynzEP7EnPsi