PECOS v1.2.5

Enhancements

• Update PyTorch dependency to >= 2.0 and fix sparse gradient clipping by @jiong-zhang in #288

Bug Fixes

• Bump requests from 2.31.0 to 2.32.0 in /examples/xr-transformer-neurips21 by @dependabot in #287
• Bump tqdm from 4.42.1 to 4.66.3 in /examples/qp2q by @dependabot in #285
• Bump tqdm from 4.62.3 to 4.66.3 in /examples/xr-transformer-neurips21 by @dependabot in #286

Full Changelog: v1.2.4...v1.2.5

PECOS v1.2.4

Enhancements

• Implement Memory-mapped MLModel by @OctoberChang in #275

Bug Fixes

• Generate directory if path not exist when saving a XLinearModel. by @jianhao2016 in #281

New Contributors

• @jianhao2016 made their first contribution in #281

Full Changelog: v1.2.3...v1.2.4

PECOS v1.2.3

Enhancements

• Add XLinear mmap unittest by @weiliw-amz in #274
• Robust stop criteria for fitting Platt transform by @jiong-zhang in #277

Bug Fixes

• ci: Use GITHUB_OUTPUT envvar instead of set-output command by @arunsathiya in #276
• Fix broken link in documentation by @justindhillon in #278

New Contributors

• @arunsathiya made their first contribution in #276
• @justindhillon made their first contribution in #278

Full Changelog: v1.2.2...v1.2.3

PECOS v1.2.2

Enhancements

• Add C implementation to fit Platt scaling by @jiong-zhang in #266
• Implement C/C++ PairwiseANN w/ Python API by @OctoberChang in #268
• Experiment Code of PEFA Paper for WSDM24 by @OctoberChang in #269

Bug Fixes

• move platt-transform exception to python by @jiong-zhang in #270
• Fix train/predict bugs in PairwiseANN by @OctoberChang in #271
• Add target clipping and update init point for platt transform by @jiong-zhang in #272

Full Changelog: v1.2.1...v1.2.2

PECOS v1.2.1

Enhancements

• Enhance mmap ValStore efficiency by @weiliw-amz in #262 #264

Bug Fixes

• Fix the missing val_prealloc reallocation during dynamic preallocation for mmap value store by @hallogameboy in #261

Full Changelog: v1.2.0...v1.2.1

PECOS v1.2.0

Highlights

• Memory-mappable 2D value store (float32 values or byte strings) for out-of-core database and feature stores.
• Batch lookup for memory-mappable hashmap and value store.

Enhancements

• Memory-mappable 2D value store for float32 values and byte strings by @weiliw-amz and @hallogameboy in #257 #260
• Batch lookup for memory-mappable hashmap by @weiliw-amz and @hallogameboy in #256 #258

Bug Fixes

• Bump certifi from 2021.5.30 to 2023.7.22 in /examples/xr-transformer-neurips21 by @dependabot in #251

Full Changelog: v1.1.0...v1.2.0

PECOS v1.1.0

Highlights

• Extend Sparse Inner Products to support DRM2DCM.
• Memory-mappable Hashmap (int-to-int, str-to-int) for fast, out-of-core retrieval of large key-value store.

Enhancements

• Extend Sparse Inner Products to support DRM2DCM by @OctoberChang in #220
• Int2Int and Str2Int memory-mappable Hashmap based on Ankerl Hashmap by @weiliw-amz in #222

Bug Fixes

• Update setup.py according to dependences upgrades by @weiliw-amz in #215
• Bump requests from 2.26.0 to 2.31.0 in /examples/xr-transformer-neurips21 by @dependabot in #224
• Bump scipy from 1.7.1 to 1.10.0 in /examples/xr-transformer-neurips21 by @dependabot in #243
• Bump scipy from 1.4.1 to 1.10.0 in /examples/qp2q by @dependabot in #242

Documentations

• Update Publication List as of May 2023 by @OctoberChang in #216
• Add PINA example folder by @jiong-zhang in #225
• feat: add FM project by @jybai in #211

New Contributors

• @jybai made their first contribution in #211

Full Changelog: v1.0.0...v1.1.0

PECOS v1.0.0

Highlights

• Enable C++ interface of HNSW with 4bit PQ for faster inference of ANN
• Release AWS Batch CDK to automate AWS infra construction for distributed training
• Accelerate PECOS Hierarchical Clustering
• Memory-mapped Utility of XR-Linear and HNSW for faster loading large models

Enhancements

C++ Interface of HNSW with 4bit Product Quantization

• Add HNSW-QG AVX512f version by @Patrick-H-Chen in #175
• Add HNSWPQ4Bits example by @Patrick-H-Chen in #176

CDK for Automated Distributed XR-Linear Training

• AWS multi-node Batch CDK for distributed PECOS by @weiliw-amz in #178
• Fix AWS multi-node batch cdk docker file bug by @weiliw-amz in #197
• AWS CDK build image from local PECOS source by @weiliw-amz in #198
• Add make clean in CDK Dockerfile by @weiliw-amz in #201
• Add JSON API for distributed XR-Linear training by @jiong-zhang in #208

Accelerated Hierarchical Clustering

• Accelerate clustering with sparse-dense vector and parallel sorting by @yaushian in #183
• Implement sampling in clustering.hpp by @yaushian in #187
• Python interface for sampling of hierarchical clustering by @yaushian in #190
• Switch between sdvec and dense vec by @yaushian in #203
• Python code for sampling of distributed clustering by @yaushian in #204

Memory-mapped Utility for XR-Linear and HNSW

• Refactor chunked matrix for accommodating memory-mapped utility module by @weiliw-amz in #192
• Add memory-mapped utilility module by @weiliw-amz in #189
• Memory-mapped XLinear Model by @weiliw-amz in #195
• Enabled C++ memory-map usage in PECOS-HNSW by @OctoberChang in #209
• Enable Python API of memory-map IO for PECOS-HNSW by @OctoberChang in #212

Bug Fixes

• Temporarily Pin setuptools<=60.0 in Wheel Build by @weiliw-amz in #173
• Remove PretrainedTransformer Vectorizer to avoid Pytest Error by @OctoberChang in #179
• Bump joblib from 1.0.1 to 1.2.0 in /examples/xr-transformer-neurips21 by @dependabot in #181
• Pin build-pypi-wheel runner image to ubuntu-20.04 by @weiliw-amz in #188
• Remove deprecated hierarchical_kmeans and related unittest by @weiliw-amz in #193
• Remove Ubuntu-18.04, Python3.6 and add Ubuntu-22.04, Python3.10 in Actions by @weiliw-amz in #194
• Remove sphinx_rtd_theme from requirements by @weiliw-amz in #200
• Upgrade codeql action to v2 by @weiliw-amz in #199
• Bump Python and Ubuntu version in README.md by @weiliw-amz in #213

Documentations

• Refactor PECOS KDD2022 tutorial instructions and materials by @hallogameboy, @jiong-zhang, and @OctoberChang in #164 #165 #167 #168 #170
• Update PECOS-related paper citations by @jiong-zhang and @OctoberChang in #169 #180
• Add the example folder of PINA by @jiong-zhang in #182

Full Changelog: v0.4.0...v1.0.0

PECOS v0.4.0

Highlights

• Enable distributed XR-Transformer fine-tuning
• Enable the capability of large-batch prediction for ANN HNSW
• Release interactive hands-on tutorial materials

Enhancements

• Unit test for sorted_csc, sorted_csr by @chepingt in #139
• Unit test for csr_row_softmax by @houyuhan98 in #141
• Bump numpy from 1.21.0 to 1.22.0 by @dependabot in #145 #146
• Release the materials for the PECOS hands-on tutorial in KDD 2022 by @hallogameboy in #153 #154 #161
• Enable the capability of large-batch prediction for HNSW by @OctoberChang in #156
• Distributed XR-Transformer fine-tuning by @jiong-zhang in #144 #162

Bug Fixes

• Fix argument-passing issue in smat_util.sorted_csc by @jiong-zhang in #134
• Fix indptr overflow issue in block_diag_csr() by @OctoberChang in #136
• Fix the yum group install command in README by @hallogameboy in #138
• Change file names for windows compatibility by @YangyiLi001 in #143
• Avoid triggering CodeQL on push for Dependabot branches by @weiliw-amz in #148
• Fix Pypi release version error by @weiliw-amz in #163

Deprecation

• Deprecate imbalanced hierarchical K-means from clustering and semantic indexing by @hallogameboy in #151

New Contributors

• @chepingt made their first contribution in #139
• @houyuhan98 made their first contribution in #141
• @YangyiLi001 made their first contribution in #143
• @xiusic made their first contribution in #147

Full Changelog: v0.3.0...v0.4.0

PECOS v0.3.0

Highlights

• Enable distributed training for XLinear
• Enable PECOS for aarch64(arm64) CPU Architecture
• Enhance pecos.ann.hnsw with Function Multi-Versioning (FMV) technique to automatically select the best supported SIMD instructions (SSE, AVX2, AVX512) at runtime
• Reduce CPU memory usage in pecos.xmc.xtransformer training

Enhancements

• Add distilbert model. by @mo-fu in #97
• add CNAME by @jiong-zhang in #104
• Bump numpy from 1.20.3 to 1.21.0 in /examples/qp2q by @dependabot in #110
• enable Function Multi-Versioning (FMV) to support AVX512 by @rofuyu in #111
• Modify supported Python version by @weiliw-amz in #113
• Enabling PECOS for aarch64(arm64) CPU Architecture by @weiliw-amz in #114
• Update OpenBLAS Version for x86 Wheel Build by @weiliw-amz in #117
• SIMD Functions for aarch64(ARM64) by @weiliw-amz in #115
• Add profile_util module by @weiliw-amz in #121
• Fix FMV setup link flag and add test wheel CI by @weiliw-amz in #119
• Fix xlinear.reconstruct_model; Add PII embedding by @weiliw-amz in #120
• Add Distributed PECOS XLinear Modules by @weiliw-amz in #123
• Add distributed PECOS README by @weiliw-amz in #127
• update HNSW README and save/load in Python API by @OctoberChang in #129
• Improve XR-Transformer memory efficiency by @jiong-zhang in #128

Bug Fixes

• properly set Text2Text prediction argument by @OctoberChang in #101
• Fix HiearchicalMLModel pred-params initialization and add bugs by @weiliw-amz in #103
• minor bug fix in XR-Transformer exp script by @jiong-zhang in #106
• fixed multithreading bugs in py hierarchical kmeans by @OctoberChang in #108
• set pytest of hierarchical kmeans with single thread by @OctoberChang in #109
• Fix relative path in distributed README by @weiliw-amz in #130

Experiment Codes for Publications

• add overlap-clustering (Liu et al.) in NeurIPS21 by @xuanqing94 in #98
• add MACLR codes by @xyh97 in #100
• update experiment code for pecos jmlr paper by @OctoberChang in #107
• update Philip's experiment code into example folder by @OctoberChang in #118

New Contributors

• @mo-fu made their first contribution in #97
• @xuanqing94 made their first contribution in #98
• @xyh97 made their first contribution in #100
• @dependabot made their first contribution in #110

Full Changelog: v0.2.3...v0.3.0