CHANGELOG.md

Changelog

All notable changes to this project will be documented in this file.

[unreleased]

⚙️ Miscellaneous Tasks

• Enhance CHANGELOG update process with branch detection and push handling

[0.1.0] - 2026-02-22

🚀 Features

• Make function to download pretrained models public and update function name
• Refactor bidirectional LSTM implementation for improved performance with batch sizing and readability
• Fix CUDA error in LSTM weights by temporarily using CPU device for forward pass
• Add examples for loading and fine-tuning alphapeptdeep models
• Add GBDT Classifier to Redeem project
• Add pi0 estimation to P-P plot in redeem-classifiers crate
• Update ReDeeM README with usage instructions and note about crates availability
• Update plot functions in redeem-classifiers crate
• Add x-axis title to score histogram plot in redeem-classifiers crate
• Update pi0 estimation in P-P plot for redeem-classifiers crate
• Add DataTables and FileSaver.js for interactivity and downloading in report module
• Update DataTables and FileSaver.js dependencies for interactivity and downloading in report module
• Add JavaScript for tabs in report module
• Add box plot function to redeem-classifiers crate
• Update plot_score_histogram function in redeem-classifiers crate
• Add legend orientation to scatter plot in redeem-classifiers crate
• Improve table container
• Add colResize feature to DataTables in report module
• Update DataTables colResize CSS link in report module
• Add web_gl_mode for large data in scatter plot
• Update legend orientation in scatter plot to vertical
• Generate random plot IDs for each plot in the report
• Add installation instructions for XGBoost library in README.md
• Fix typo in log message for Input26aaModPositionalEncoding::forward method
• Add new modules for training and loading data in redeem-cli
• Add Dockerfile for CUDA-based application containerization
• Add inference functionality to redeem-cli
• Implement custom error handling for TDC calculations
• Add error handling for Experiment initialization and validation
• Add redeem-properties-py PyO3 bindings package
• Add from_pretrained constructor to RTModel, CCSModel, MS2Model
• Accept modified peptides directly in predict; auto-parse mods
• Return fragment annotations alongside MS2 intensities in predict
• Annotate CCS predict output with charge state per peptide
• Add predict_df methods returning pandas/polars DataFrames
• Add pretrained feature for redeem-properties integration
• Add dynamic layer construction for DecoderLinear from VarMap
• Infer hyperparameters dynamically for CCS and RT models' decoders
• Add hyperparameter inference for CNN-Transformer models
• Update model constructors to accept optional constants path
• Make constants_path optional in CCSModel and MS2Model constructors
• Add functions to locate and validate pretrained models in Python
• Add optional output denormalization for model predictions
• Add CCSCNN model prediction example with detailed output
• Enhance model name mapping in from_str for pretrained models
• Add model architecture retrieval and detailed summaries for CCS, MS2, and RT models
• Add parameter count and summary methods for RT, CCS, and MS2 models
• Add multiplier and exclude_zeros parameters to MS2Model predictions
• Add PropertyPrediction class for unified peptide property predictions
• Add rustyms dependency and mz_utils module to enhance functionality
• Add m/z computation functions and enhance MS2Model and PropertyPrediction for m/z annotation
• Add unified prediction functionality with PropertyPrediction for combined RT, CCS, and MS2 outputs
• Enhance CCS and MS2 models with input expansion and ion mobility annotation
• Enhance MS2Model with precursor m/z annotation and update Rust model wrappers
• Add comprehensive tests for RT, CCS, and MS2 models with DataFrame support and annotations
• Implement utility functions for computing precursor and product m/z values from ProForma notation
• Update Python wheel and source distribution workflows to include pretrained features
• Add Sphinx documentation structure with Makefile, requirements, and initial content
• Add custom CSS and logo for Sphinx documentation theme
• Add new logo images for dark theme in documentation
• Add README.md for Sphinx documentation setup and usage instructions
• Add GitHub Actions workflow for building and uploading Sphinx documentation
• Add Read the Docs configuration for building documentation
• Implement Array2 and Array1 structures with associated methods for matrix and vector operations
• Enhance SVM example to accept CSV path and column count from CLI arguments
• Add reporting functionality to SVM and XGBoost examples, including score distribution plots
• Add preprocessing utilities and integrate scaling and normalization options in SemiSupervisedLearner
• Update dependencies and improve prediction output formatting in examples and models
• Add report-builder dependency and re-export Report and ReportSection
• Add pretrained model support and enhance CLI options for inference
• Implement Percolator .pin TSV reader and related utilities
• Add scoring functionality for Percolator .pin files with semi-supervised classifier
• Introduce RankGrouping for enhanced rank inference and deduplication in scoring
• Add feature_weights method to ClassifierModel and implement in SVMClassifier for linear feature importances
• Add HTML report generation for scoring results with histograms
• Add support for SVM and XGBoost models in CLI with report generation toggle
• Add warmup_fraction parameter for learning rate scheduling in training functions
• Update logo images
• Implement pretrained model loading and caching functionality
• Enhance PyPI publishing workflow with environment and permissions
• Add BSD 3-Clause License file
• Update pyproject.toml with additional metadata and classifiers
• Update version to 0.1.0 in pyproject.toml
• Add workflow to generate changelog and release notes
• Add README.md for redeem-properties library with installation and usage instructions
• Add README.md for redeem-classifiers library with detailed features, installation instructions, and usage examples
• Add GitHub Actions workflow for releasing CLI binaries
• Add comprehensive tests for redeem_properties functionality
• Add Docker publish workflow and update Dockerfile for multi-stage build
• Update CLI to allow optional config paths and print templates when not provided
• Add download_pretrained_models function and update tests for its existence

🐛 Bug Fixes

• Remove linux specific and unused imports for bilstm
• Fix loss computation for rt cnn lstm model during fine tuning.
• Correct loss computation for rt cnn lstm model during fine tuning
• Put bert weights on CPU since layerNorm is not implemented for CUDA
• Squeeze instrument indices tensor
• Remove instrument_indice squeezing
• Modication name and indice retrieval
• Correct display formatting for ExperimentError single class message
• Ensure version is correctly specified in pyproject.toml
• Downgrade ndarray version to 0.15 to match numpy's dependency
• Rename redeem_properties_py function to _lib for consistency
• Update model and constants paths to use pretrained models directory
• Update CCS and MS2 prediction parameters for consistency and clarity
• Update light theme logo path in Sphinx configuration
• Correct light-svm dependency version in Cargo.toml and add clarification comment
• Update prediction conversion to f64 in GBDT example and enhance SVM scorer error handling
• Correct URL formatting for pretrained models
• Correct path formatting for pretrained models and add unit test for model download
• Update error handling for model type parsing in handle_classifiers function
• Update plotly usage for histogram layout and improve label handling in scoring functions
• Update tqdm dependency to version 0.8.0
• Update logo image sources for light and dark color schemes in README
• Update workflow to trigger on both master and develop branches
• Update Python installation instructions and import paths in README
• Remove optional embedding of pretrained model files from Cargo.toml
• Update tqdm dependency to use specific git branch
• Add error module and handle Experiment creation errors
• Remove ndarray dependency from Cargo.toml
• Update pretrained model string matching for consistency
• Update pretrained model resolution to use architecture method
• Update Read the Docs configuration for Python and Maturin installation
• Update Rust version to 1.86 in Read the Docs configuration
• Update Rust version to 1.91 in Read the Docs configuration
• Update Read the Docs configuration to omit pretrained feature during build
• Update TODO list to save target normalization min/max values in safetensor
• Clean up .gitignore by removing unnecessary comments and ensuring proper formatting
• Update README to correct terminology from "rescoring" to "scoring" for classifiers
• Update README links to point to correct README files for each crate
• Update Dockerfile to copy Cargo.lock for dependency management
• Correct project name format in pyproject.toml

💼 Other

• Fine tuning of rt cnn-lstm model and add tqdm logging
• Progress logger to dynamically update desc
• Cargo to use feature branch of tqdm for dynamic desc update
• Peptide data structure to hold peptide information and optional peptide property outputs
• Data handling module
• PeptideData structure to models and add fine-tuning method for CCS
• MS2 fragment intensities to PeptideData structure and add fine-tuning for MS2 bert model
• Method to download pretrained models
• Method get_modification_string to retrieve modifications from modification tsv
• Update num_batches for progress logger
• TODOs
• Todos
• AtomicF64 import
• Update dependencies
• Trace messages during SVM fitting
• Readme
• Method to load tensors from pytorch pth files or safetensor files
• MesaTeE GBDT classifier to SemiSupervisedLearner
• Examples
• Scatter plot
• Data table in report
• Plot auto sizing based on window
• Properties cargo dependencies for candle
• TransformerEncoder and SeqTransformer block
• Encoder26aaModChargeCnnTransformerAttnSum implementation
• Readme
• RT Norm struct to set type of normalization
• Modification.tsv asset
• Bilstm forward with state
• Type annotation for bilstm
• Update Python bindings installation to remove unnecessary features flag
• Remove unnecessary features flag from wheel build command
• Update reqwest dependency to disable default features and enable rustls-tls
• Update macOS target versions to use 'macos-latest'

🚜 Refactor

• Update device handling in bilstm.rs and building_blocks.rs
• Update device handling in bilstm.rs and building_blocks.rs
• Update fine-tuning method to include batch size parameter, refactor rt model fine tuning with batch sizing
• Update fine-tuning method to include batch size parameter and improve ccs model fine-tuning with batch sizing
• Update fine-tuning method to include batch size parameter and improve ms2 model fine-tuning with batch sizing
• Move get_asciii_indices to featuize
• Remove unused import in bilstm.rs file
• Update featurize.rs to remove unused import and align with recent changes
• Refactor PredictionResult and PredictionValue types
• Add ion mobility to CCS conversion
• Fine-tuning log message
• Update fine-tuning log message for model training
• Update ModelParams and ModelType
• Add SVM classifier to redeem-classifiers crate
• Remove rayon parallel processing for fine-tuning
• Update forward pass in MS2BertModel
• Update LSTM forward pass in Encoder26aaModCnnLstmAttnSum and Encoder26aaModChargeCnnLstmAttnSum
• Add CUDA support for candle-nn and candle-transformers crates
• Update ModelParams and ModelType
• Remove commented out code in ModelType enum
• Update feature scores log message in log_input_data_summary
• Remove unused code and update cross-validation log messages
• Add SVM and XGBoost semi-supervised learning examples
• Add class percentage balance options for training in SemiSupervisedLearner
• Add log message for training data selection in SemiSupervisedLearner
• Update log message for training data selection in SemiSupervisedLearner
• Comment out remove unlabelled psms
• Update loss type handling in GBDTClassifier
• Update loss type handling in GBDTClassifier
• Comment out unused code in psm_scorer.rs
• Update log message for training data selection in SemiSupervisedLearner
• Remove redundant sentence in redeem-properties crate's README.md
• Update model structs to use 'static lifetime for VarBuilder
• Update model structs to use 'static lifetime for VarBuilder
• Update ModelClone trait to include Send and Sync bounds
• Update DLModels struct to remove unnecessary Arc and Mutex wrappers for model fields
• Update peptide modification handling to support mass shifts and UniMod annotations
• Peptide encoding
• Bilstm
• Optimize peptide sequence featurization and one-hot encoding
• Update RTCNNLSTMModel forward method to improve performance and readability
• Update redeem-properties crate models to remove unused imports and improve code organization
• Update RTCNNLSTMModel forward method to improve performance and readability
• Add RT-CNN Transformer model and update redeem-properties crate models
• Add early stopping to property training
• Remove unnecessary cargo update command in Dockerfile
• Update dependencies and descriptions in Cargo.toml files
• Update redeem-properties crate models and add new modules for training and loading data in redeem-cli
• Add new modules for training and loading data in redeem-cli
• Update Dockerfile to optimize build process and clean up artifacts
• Add CCSCNNTFModel implementation
• Update RTCNNTFModel implementation and remove unused code
• Improve regex pattern for extracting modification indices in peptdeep_utils.rs
• Update redeem-properties crate models for CCS prediction
• Add new fields to load_peptide_data function in redeem-cli
• Add stats module to redeem-properties crate, and add lr scheduler
• Add precursor mass field to PeptideData struct in redeem-properties crate
• Add plot_training_metric function to redeem-cli crate
• Update early stopping logic in ModelInterface implementation
• Update plot_losses function in redeem-cli crate
• Update config loading logic in redeem-cli crate
• Update RT-CNN-LSTM and RT-CNN-Transformer models in redeem-properties crate
• Update hidden_dim and decoder size in CCSCNNTFModel
• Clean up trace comments
• Update peptide data loading logic in redeem-cli crate
• Update PeptideData struct to use u8 for string fields
• Update mod_to_feature loading to use Arc for key in RTCNNTFModel and CCSCNNTFModel
• Improve error handling in redeem-cli crate
• Optimize contiguous operations in building_blocks.rs
• Update rank feature based on new classifier scores
• Update examples in classifiers crate
• Update rank feature and log rank changes in Experiment class
• Set log level to debug in main function
• Update loading of modifications to use byte slice instead of file path
• Update data handling to use TargetNormalization instead of RTNormalization
• Add once_cell dependency for redeem-properties crate
• Update normalization field in Redeem CLI properties
• Update loading of modifications to use byte slice instead of file path
• Update training configuration in redeem-properties crate
• Update AAEmbedding constructor signature to accept VarBuilder instead of Device
• Update semi-supervised learning to return updated ranks along with predictions
• Update Redeem CLI to use RTCNNTFModel for inference
• Update data handling to extract "rank" feature column as 1D array of u32s
• Improve bidirectional LSTM input handling for contiguous tensors
• Improve bidirectional LSTM input handling for contiguous tensors
• Improve initialization of hidden states in BidirectionalLSTM
• Improve bidirectional LSTM forward and backward processing
• Clone contiguous tensor in BidirectionalLSTM for improved handling
• Improve logging in BidirectionalLSTM backward processing
• Update Dockerfile to separate build and runtime stages, improve CUDA handling
• Use pretrained module registry for from_pretrained constructors
• Enhance documentation for locate_pretrained function to clarify model path resolution
• Replace ModelParams with ModelConfig across classifiers and update related implementations
• Replace SemiSupervisedModel with ClassifierModel trait and update related implementations
• Remove unused utils module from model implementations
• Clean up imports and enhance SemiSupervisedLearner with max_iterations parameter
• Enhance SVM and XGBoost classifiers with improved parameter handling and logging for fallback scores
• Reorganize module imports for clarity and consistency across building_blocks, lib, models, and utils
• Improve code readability and structure in BidirectionalLSTM implementation
• Simplify DecoderLinear and DecoderMLP implementations with improved error handling and logging
• Remove unused projection weight field from Encoder26aaModCnnTransformerAttnSum and Encoder26aaModChargeCnnTransformerAttnSum
• Improve code organization and readability in featurize.rs with consistent formatting and sequential accumulation
• Clean up imports and improve formatting in nn.rs for better readability
• Adjust training mode default to false in RTCNNLSTMModel and update dropout application in CCSCNNLSTMModel
• Add TargetNormalization and train_var_prefixes parameters to training and fine-tuning methods in CCS, MS2, and RT model wrappers
• Enhance CCSCNNTFModel and RTCNNTFModel with flexible decoder head selection and improved constructor methods
• Simplify CNN output handling in Encoder26aaModCnnTransformerAttnSum and Encoder26aaModChargeCnnTransformerAttnSum by removing unnecessary pre-projection copy
• Clean up RTCNNTFModel constructor by removing unnecessary comments and improving clarity in varmap handling
• Clean up whitespace and improve formatting in TargetNormalization and PeptideData structures
• Reorganize imports and clean up whitespace in logging module
• Reorganize imports and improve formatting in peptdeep_utils.rs
• Enhance TrainingStepMetrics by adding MAE, RMSE, and R² calculations and updating metric summaries
• Reorganize imports and improve formatting in utils.rs
• Consolidate import statements and improve formatting in alphapeptdeep examples
• Improve encoder variable scoping and set training mode in CCSCNNTFModel
• Reorganize imports and improve logging format in main.rs
• Add candle-nn dependency to Cargo.toml
• Improve formatting and organization of code in util.rs
• Simplify header index lookup and improve normalization logic in load_peptide_data
• Clean up imports and enhance PropertyTrainConfig with additional normalization and head type fields
• Reorder module declarations and improve sample_indices function documentation
• Enhance write_peptide_data function to include original targets and improve header handling
• Reorder imports in xgboost.rs for improved organization
• Improve MS2 target tensor construction with enhanced error handling and logging
• Update prediction and target extraction to use flatten_all for improved tensor handling
• Update pretrained models URL and paths for new pretrained models added
• Move individual crates out of crate folder into main root project
• Remove test_basic.py as part of codebase cleanup
• Reorganize CI workflow for Rust and Python tests

📚 Documentation

• Update title in index.md for clarity and consistency
• Add README with examples for redeem-classifiers usage
• Enhance module documentation across the crate with detailed descriptions
• Update README.md to enhance project description and usage examples
• Add closing code block for Unified prediction section in README
• Update README to emphasize CLI is still under development
• Update examples to include necessary imports and correct assertions
• Add badges for Rust, PyPI, and documentation in README
• Update getting started guide to clarify pretrained model downloads

🧪 Testing

• Ensure pretrained models are downloaded before tests that require them
• Ensure pretrained models are loaded before running tests
• Import necessary modules for testing SemiSupervisedLearner
• Set up virtual environment for Python dependencies and tests

⚙️ Miscellaneous Tasks

• Update dependencies and add new crates
• Add report module and plots module to redeem-classifiers crate
• Update env_logger dependency to version 0.8.4
• Update xgboost dependency to use custom repository and branch
• Update XGBoost and Linfa dependencies for SVM classifier support as optional features
• Update xgboost dependency to use custom repository and branch that is more up-to-date
• Update model type handling for XGBoost and SVM classifiers for optional feature dependency
• Update xgboost dependency to use custom repository and branch for more up-to-date version
• Update Xxgboost test
• Remove ndarray_stats
• Change rust-xgboost path back to git url
• Add trace logs for one hot encoded data and input tensor devices
• Refactor one_hot method for improved readability and error handling
• Update instrument_indices logging to use f64 type
• Update instrument_indices logging to use f32 type
• Improve logging and device handling in MetaEmbedding::forward method
• Update logging for charges and nces in MetaEmbedding::forward method
• Add debugging trace logs for debugging shape/device issues
• Update charge variable type to integer in peptdeep_utils.rs
• Update AAEmbedding with trace logs
• Update in MS2BertModel::forward method
• Add trace logs for debugging shape/device issues in AAEmbedding and Input26aaModPositionalEncoding
• Add trace logs for encoding peptide sequence and tensor shape in ModelInterface::encode_peptide method
• Refactor get_aa_indices to filter out invalid amino acid characters
• Update gbdt dependency to use custom branch for decision function support
• Update GBDTClassifier to use decision_function for predictions
• Update logging level for invalid amino acid characters in get_aa_indices method
• Update XGBoost parameters for semi-supervised learning
• Update XGBoostClassifier evaluation matrix label to "eval"
• Update XGBoost parameters for semi-supervised learning example
• Update XGBoostClassifier to use auc for eval early stopping
• Add Clone trait implementation for ModelInterface
• Update dependencies in redeem-properties crate
• Comment out debug print statements in Encoder26aaModCnnLstmAttnSum
• Add concurrency settings to GitHub Actions workflows for docs and Rust
• Update Docker publish workflow to trigger on release events only

◀️ Revert

• Apply_bidirectional_layber and forward_wtih_state in bilstm to earlier version before refatoring

[0.1.0-alpha] - 2025-02-06

🐛 Bug Fixes

• Forgotten commits, and getting ms2 prediction model forward method to work
• Process intensity predictions to output
• Use var_store ref in ccs encoder
• Make aa_size from constant yaml optional
• Ccs decoder dimensions
• Some bugs with xgboost and change some parameters for objective method
• Add log dependency to cargo

💼 Other

• Initial commit of refactoring and abstraction from sage impl
• README
• Cargo toml
• Fine tuning method
• Bulding blocks for ms2 bert model
• Peptdeep ccs prediction model
• Test for loading from pytorch pth model
• README
• Peptide encoding to CCS model
• Prediction method, and fix bugs in SeqAttn
• Main ccs model interface
• Ms2 model interface
• Todo
• Todo
• Todo
• Print statements
• Readme
• Get device from string
• Print statements
• Model_interface
• DLModel structure to hold rt, im and ms2 models
• Ccs to mobility util:
• SelectKBest univariate feature selection with F-score regression
• TODO
• Experiment struct for holding PSM score data for semi-supervised learning
• Xgboost classfiier model
• Target-decoy comp qval estimation
• Semi-supervised learner and xgboost model
• Lib parts
• TODO
• Readme
• Minor
• Minor
• Logo
• Readme
• Param struct to DLModels to so set instrument and nce values
• Nce to f32
• Rust version to stable rust release on compute canada
• Model params struct to control model param flexibility
• Serialize/deserialize ModelParams
• Serde improt and depend
• Clone to ModelParams
• Cargo to huggingface candle and add cuda feature
• Device type retrieval
• Cargo add cuda as a feature to avoid failure on machines without GPU
• Add tqdm logging interface
• Tqdm and rayon to dependency
• Sysinfo for getting resident set size

🚜 Refactor

• Move property prediction into a separate crate
• Add properties prediction create and add more modules to building block
• Peptdeep RT model

⚙️ Miscellaneous Tasks

• Cleanup pept deep utils