Fix/senkin fit walltime - #70
Merged
Merged
Conversation
New arguments --lgbm_learning_rate, --lgbm_n_folds, --lgbm_max_bin and --lgbm_n_jobs (worker processes over protein targets, from the library's train_lightgbm_kfold). The network fold count (--n_folds) is now separate from the LightGBM one. Defaults unchanged for now. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
…placed by main after the library release) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
…ed with --no-deps) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
…rker budget Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
learning rate 0.1, at most 200 rounds with early stopping 20, 3 LightGBM folds (networks keep 5), 63 histogram bins, one worker per CPU within the allocated memory. Measured on the NeurIPS 2021/2022 CITE datasets: the LightGBM stage keeps its per-target quality within 0.003 of a 1000-round run, and the final predictions are unchanged compared with 300 rounds and 5 folds (2021: Pearson per cell 0.822 either way). Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
…n the library Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
200 rounds barely cut the time on the 2022 data (early stopping already ends most models around 170 rounds); 100 rounds keep the final predictions unchanged and leave a safe margin under the 8 h limit. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
benjaminfreyuu
approved these changes
Sep 21, 2026
benjaminfreyuu
left a comment
benjaminfreyuu
left a comment
Contributor
There was a problem hiding this comment.
I trust that the stated maintenance of performance/results is true. Then approve
| Maximum LightGBM boosting rounds (early stopping applies). The original solution used up to | ||
| 10000 rounds at learning rate 0.01, which takes days on the benchmark datasets. At learning | ||
| rate 0.1, 100 rounds keep the LightGBM per-target quality within 0.015 (NeurIPS 2021) / 0.003 | ||
| (NeurIPS 2022) of 300 rounds, and the final predictions of the networks are unchanged. |
| "n_folds": 5, | ||
| "lgbm_boost_rounds": 10000, | ||
| "lgbm_early_stopping": 100, | ||
| "lgbm_n_folds": 3, |
Contributor
There was a problem hiding this comment.
why decrease nr of folds? Also due to time constraints?
Contributor
Author
There was a problem hiding this comment.
Yes, exactly
Contributor
Author
There was a problem hiding this comment.
Results were comparable anyway
| "nn_epochs": 100, | ||
| "n_tsvd_components": 100, | ||
| "lgbm_learning_rate": 0.1, | ||
| "lgbm_max_bin": 63, |
Contributor
There was a problem hiding this comment.
does this param. keep results unchanged? It is stated as such above, so trusting this
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
2 participants
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.
Improve senkin_tmp model scalability
In the previous version, senkin_tmp model largely failed because of the time limit. This PR reduces the number of epochs and increases learning rate to train the models faster
Checklist before requesting a review
I have performed a self-review of my code
Check the correct box. Does this PR contain:
Proposed changes are described in the CHANGELOG.md
CI Tests succeed and look good!