Conversation
goal is to be able to run a single script that will update all the model figures (both the interactive version for the leaderboard and the PDF version for the paper)
…ze_element_errors.py to run former as part of update_model_figs.py
janosh
commented
Jul 26, 2023
| except ImportError as exc: | ||
| raise ImportError( | ||
| "pdfkit not installed\nrun pip install pdfkit && brew install " | ||
| "homebrew/cask/wkhtmltopdf\n(brew is macOS only, use apt on linux)" |
Owner
Author
There was a problem hiding this comment.
Reason: wkhtmltopdf requires Rosetta, i.e. is not natively M1 Mac compatible.
…ts/analyze_element_errors.py'
janosh
added a commit
that referenced
this pull request
Jul 31, 2026
* add scripts/model_figs/update_all_model_figs.py goal is to be able to run a single script that will update all the model figures (both the interactive version for the leaderboard and the PDF version for the paper) * prepare MACE submission (readme.md, metadata.yml, test_mace.py) * fix model name still set to chgnet in test_mace.py * wandb collect all dep versions in single dict * delete matbench_discovery.DEBUG global * roc_prc_curves_models.py fix n_rows x n_cols in out filename * test_mace.py add relax trajectory recording * add MACE + ALIGNN checkpoints figshare urls to class DataFiles * rename BOWSR + MEGnet -> BOWSR * ensure out_path matches glob_pattern in join scripts * load_df_wbm_with_preds() use 1st matching df column * refactor df_to_pdf() from wkhtmltopdf to weasyprint * improve Figshare description * update most figures with MACE results * extract scripts/model_figs/per_element_errors.py out of scripts/analyze_element_errors.py to run former as part of update_model_figs.py * add MACE to site/src/figs/model-run-times-bar.svelte * revert 'extract scripts/model_figs/per_element_errors.py out of scripts/analyze_element_errors.py' * add weasyprint to df-pdf-export optional deps * fix test_df_metrics failing from MACE R^2 of -1.291 being below -0.9 * fix df_to_pdf if weasyprint not installed
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
1 participant
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.
As released in NeuRIPS paper and further developed in 2022 preprint.
This submission uses the
2023-07-14-mace-universal-2-big-128-6.modelcheckpoint pre-trained by the MACE authors on the original M3GNet training set.MACE relaxed each test set structure using
ase.optimizers.FIREuntil the maximum force in the training set drops below 0.05 eV/Å or 500 optimization steps are reached, whichever occurs first.