Make the FlexCNN multitask model the default (v4.1.1) - #115
Merged
Conversation
The fused-trunk multitask model trained across 6,543 LC setups, bundled since 4.1.0 as an opt-in, is now what every core function and the command line load when no model is given. The 4.0 default stays bundled as `deeplc.core.LEGACY_MULTITASK_MODEL` so existing workflows can pin it. An uncalibrated `predict()` on a multitask model that carries setup names now reports the setup named by `DEFAULT_TASK_NAME` (`PXD005573_mcp`, the 200-minute gradient the DeepLC 1.x to 3.x models were trained on) rather than head 0, which for the new default was whichever setup sorted first. Calibration, fine-tuning and `return_matrix=True` are unchanged. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
DEFAULT_MODELnow points atmultitask_flexcnn_model.pt(fused trunk, low-rank multitask head, 6,543 LC setups; bundled since 4.1.0 asFLEXCNN_MULTITASK_MODEL). Every core function and the CLI use it when nomodelis given.multitask_model.ptstays bundled asdeeplc.core.LEGACY_MULTITASK_MODEL; pass it asmodel=to reproduce 4.0 / 4.1.0 predictions exactly.predict()on a multitask model with setup names reportsDEFAULT_TASK_NAME = "PXD005573_mcp"(the DeepLC 1.x-3.x training setup, 200-min gradient) instead of head 0. For the new default, head 0 was an arbitrary setup (2023-06-22 - DIANN - Ajita ...).return_matrix=True, calibration and fine-tuning are unchanged.models.rst,migration.rst), README, CHANGELOG 4.1.1 entry, version bump.Verification
pytest tests: 117 passed (local, Windows, torch 2.6 CPU).test_model_opspins the legacy file explicitly;test_flexcnn::test_bundled_model_predicts_in_minutesnow checks the default-setup column; new tests intest_corecover the default constant, the default-setup contract, and that the legacy model still loads and predicts.ruff check/ruff format --check: no new findings on the changed files (repo-wide count unchanged at 78 pre-existing).Notes for release
Creating the GitHub release
v4.1.1triggerspublish.yml(PyPI, Windows installer, GHCR image).🤖 Generated with Claude Code