Initialize and refresh evaluation caches safely; migrate checkpoint keys before filtering; correct accumulation tails and distributed validation control flow and metric aggregation. Use a trainable decoder in the pretraining example and align environment installation with the pinned pip requirements. Update installation documentation. Validated with focused CPU/Gloo and RTX 4070 Ti SUPER CUDA regression checks, including same-precision real-checkpoint inference comparisons. Full downstream training accuracy and GPU multi-card execution were not benchmarked. Only the twelve reviewed source, configuration and documentation files are updated; model weights, paired model configuration and original datasets are unchanged.
Unify BioMod interfaces and centralize checkpoint compatibility
Use the bundled BioModEncoder, BioModLM and BioModLM_GO interfaces throughout. Remove obsolete CLI aliases and the external attention fallback. Centralize legacy configuration and state-dictionary translation in biomod_compat.py and document the shared asset. Model weights, paired configuration, datasets and unrelated files remain unchanged.
Remove W&B tracking and personal paths; complete BioMod naming
Remove W&B logging, account settings, CLI options and direct dependencies while preserving local logs, checkpoint saving and resume behavior. Replace historical workstation paths in two dataset summaries with repository-relative paths and refresh their manifest checksum rows. Include the previously reviewed BioMod naming changes required by the updated pretraining entrypoint; retain legacy checkpoint compatibility. Model weights, paired model configuration, original dataset payloads and unrelated files are unchanged.
Remove tests directory and refresh README references
Remove BioMod/tests/test_training_utils.py as requested. Remove the test command, links and tree entries from both project READMEs while retaining pretraining requirements. Runtime training utilities, task code, model weights, paired configuration and all data files remain unchanged.
Put dataset launch commands first in each downstream task README
Update the five task READMEs to put complete per-dataset launch commands at the beginning. Add dataset command indexes to both project READMEs. All 15 existing command blocks remain unchanged. No source code, weights, configuration, raw data, or other repository files are modified.
Unify downstream task layout and document dataset launch commands
Place each task notebook beside its README and train/evaluate entry point; organize helper code and shared module-source definitions. Add complete per-dataset launch commands and preserve declared zero-count cell annotation classes. Remove the unused environment template and stale ignore rules, and refresh project and data-directory README paths. Pretrained weights, paired configuration and raw data remain unchanged.
Remove interpretability directory and update README guides
Remove the 31 files under BioMod/interpretability as requested. Update the two README introductions, workflow references, checkpoint labels and directory trees. No external code or notebook imports depend on the removed directory. Model weights, paired configuration, raw data and all other repository files are unchanged.
Rename project-owned entrypoints, resource directories, classes and helpers to BioMod/biomod, and update imports, notebooks, documentation and command examples. Preserve old CLI flags as aliases and keep serialized state/configuration keys, raw gene-set IDs, result schemas and third-party interfaces compatible. Move the existing checkpoint/configuration pair to weight/BioMod.pth and weight/BioMod.pkl using their unchanged objects. Remove only the eight approved duplicate files from the source module-data directory and nested attention package; retain the shared copies. Raw datasets, legacy reference data copies and all weight contents remain unchanged. Validation: 90 Python 3 files, 65 notebook cells, 12 utility tests, 7 synthetic backbone comparisons and 10 classifier comparisons passed. No real model weights or datasets were deserialized, and no real-data training was run.
Make project paths portable and standardize English code comments
Resolve bundled resource defaults from script locations, preserve caller-relative path overrides, and require missing external pretraining or legacy inputs explicitly. Translate comments and docstrings to English, update command examples and documentation, and add a path helper for legacy reference code. Validation: 97 Python 3 files and 65 notebook code cells compiled; 12 existing utility tests passed; parser, example-command, resource-path and local-link checks passed. Two historical DeepCDR entrypoints retain Python 2 syntax. No training or model inference was run. This commit updates 90 source/documentation files and adds one helper; pretrained weights, paired PKL, raw datasets and other files are unchanged.
Remove new-data reconstruction evaluation utility and README references
Delete evaluate_new_data.py and its command reference from BioMod/eval_new_data. Remove the corresponding workflow, checkpoint-interface notes, and directory-tree entries from the root and source READMEs. Model weights, paired configuration, datasets, notebooks, and other source files are unchanged.
Fix downstream notebook project paths and unify tutorial links
Update the five downstream notebooks to resolve paths from the repository layout, use the bundled BioMod checkpoint paths and supported commands, and read actual task outputs. Keep training and evaluation opt-in. Standardize the main README tutorial column to Notebook links. This commit only changes these six files; model weights, paired configuration, datasets, and training source are unchanged.
18 days ago
Use BioMod model weights naming across documentation
Update eight Markdown documents to use BioMod model weights terminology. Refresh stale source-guide examples and remove historical model-size comparisons and unverified benchmark values. Actual weight filenames, paired configuration, code, and datasets are unchanged.
Remove reverse perturbation code and refresh README documentation
Remove the ten reverse perturbation implementation and protocol files. Update the project README and source README, including their repository trees. Model weights, paired configuration, datasets, and unrelated source files are unchanged.
Name model-selection splits validation in drug tasks
Use validation terminology in both drug tasks, tutorials and documentation. Add validation_label_file and validation_cell_file options with legacy aliases, and write the validation prediction field while retaining notebook support for older outputs. Keep data partitions, training behavior, dataset paths, weights and PKL files unchanged.
19 days ago
Upload BioMod with datasets and the 0810 checkpoint
Publish the BioMod source tree, model dependencies, complete downstream datasets, and the unchanged 0810 weight/configuration pair. Include the updated model implementation, shared training utilities, CPU utility tests, and English documentation. Exclude download caches, local environments, and OS metadata.
