v0.8.7

What's Changed

• [feat] add public docker images by @eric-gecheng in #537
• avoid default clear ps queue, which can cause hang in no data case by @chengmengli06 in #538
• [add docs]feat: add docs for num steps when sync_replicas: true by @lgqfhwy in #542
• [Bugfix]strip blank chars in output_cols by @eric-gecheng in #543
• Update FG document by @yangxudong in #540
• to #76994268, fix bug of BST component (default dropout rate) and CMBF model by @yangxudong in #546
• upgrade zero inflated lognormal loss by @yangxudong in #546
• support export structure path by @yangxudong in #546
• add document for jrc_loss, ziln_loss, masked_loss and so on by @yangxudong in #546

Full Changelog: v0.8.5...v0.8.7

v0.8.5

What's Changed

• Add ZILN loss for LTV prediction task [#498]
• Add xDeepFM model [#494]
• Add CIN(Compressed Interaction Network) module Add Parallel DSSM recall model [#493]
• Add NaryDisEmbedding & Transformer component [#485]
• Add AITM component & upgrade NaryDisEmbedding [#487]
• Add Highway & Attention Component And AITM model [#471]
• Support Match model configured by components [#492]
• Add TextCNN component and model [#462]
• Bug fix of SENet when run with MirroredStrategy [#443]
• Fix a few bugs

v0.7.1

What's Changed

• [feat] add docker for tf210 and py38 by @wwxxzz in #402
• [bugfix] skip label and comment, as it requires special permessions and is not necessary by @chengmengli06 in #403
• [feat] update nni docs: add resume/tensorboard by @yjjinjie in #405
• update nni docs by @yjjinjie in #406
• Nni by @yjjinjie in #407
• update nni docs by @yjjinjie in #408
• Add pdn model by @wwxxzz in #409
• add pdn to docs_recall_module by @wwxxzz in #412
• [feature]: add xflow for tf1.15 on max compute by @chengmengli06 in #304
• [docs] update easyrec processor related documents by @chengmengli06 in #416
• [doc] change_feature_pdf_to_md by @chenglongliu123 in #415
• update easyrec version to 0.7.1 by @wwxxzz in #420

Full Changelog: v0.7.0...v0.7.1

v0.7.0

What's Changed

• Features/fix docker by @wwxxzz in #395
• [feat]: place embedding on cpu if necessary by @yangxudong in #398
• [feat]: Add component-based develop Backbone Network for EasyRec by @yangxudong in #388
• [feat]: add continue train on evaluation data by @chengmengli06 in #399

Full Changelog: v0.6.4...v0.7.0

v0.6.4

What's Changed

• [feat]: add avx string split feature by @chengmengli06 in #364
• [feat]: add data link on mc tutorial inner by @wwxxzz in #393
• [feat]: add support for horovod by @chengmengli06 in #389
• [feat]: add support for multival separation using combo feature by @chengmengli06 in #396
• [feat]: update version to 0.6.4 by @chengmengli06 in #397
• [bugfix]: fix mount directory checkpoint cache problem for data science platform by @chengmengli06 in #357
• [bugfix]: Example中movieslens.config执行路径不对 by @wenyangchou in #368
• [bugfix]: fix predict out of dssm model by @wwxxzz in #370
• [bugfix]: fix predict shape of logits and probs in MIND model by @wwxxzz in #371
• [bugfix]: fix bug in match model when calculate sim() by @wwxxzz in #374
• [bugfix]: fix bugs on examples/configs/ by @wwxxzz in #376
• [bugfix]: fix fm code in layers/fm.py by @livmortis in #381
• [bugfix] fix bug of test case of fm/deepfm model by @yangxudong in #387
• [bugfix] fix finetune_checkpoint restore bug by @chengmengli06 in #386
• [bugfix]: fix sync optimizer compatibility with pai-tf by @chengmengli06 in #391
• [doc] modify project name to pai_online_project in benchmark page by @poson in #390

New Contributors

• @wenyangchou made their first contribution in #368
• @livmortis made their first contribution in #381

Full Changelog: v0.6.2...v0.6.4

v0.6.3

add support for release v0.6.3

v0.6.2a

[bugfix] add client demo missing files
[feature] add avx string split feature

v0.6.2

What's Changed

• [feat]: CSVPredictor support embedding input by @chenglongliu123 in #298

• [feat] recall sampler input support regular by @chenglongliu123 in #329

• [feat] hitrate support regular input by @chenglongliu123 in #330

• [feat] add support for sequence and key share embedding by @chengmengli06 in #331

• [feat]: add docs for ev and sample_weight by @chengmengli06 in #305

• [feat] update hpo docs and 0.1.6 support one-click-retry-multi-failed-trials by @yjjinjie in #322

• [feat] add direct cmd args for update easyrec config by @chengmengli06 in #326

• [feat] refactor input layer and update online train docs by @chengmengli06 in #320

• [feat] add combiner for raw feature by @yangxudong in #338

• [feat]: add support seq transform padding and dnn by @lgqfhwy in #344

• [feat]: add multiple ranking losses & loss weight adaptive learning & fix a few bugs by @yangxudong in #358

• [feat]: Support macOS & Windows and Support Tensorflow >= 2.12.0 by @wwxxzz in #361

• [feat]: add dice activation for dnn layer by @yangxudong in #362

• [feat]: add client demo for sending request to easyrec processor by @chengmengli06 in #363

• [feat]add udf for label by @dawn310826 in #303

• [feat]: improve hpo_tools: 1) use odpscmd 2) support kill-all-trials 3）schedule… by @yjjinjie in #350

• [feat]: update nni hpo docs by @yjjinjie in #299

• [bugfix]: fix ut bug by @chengmengli06 in #321

• [bugfix] fix typo in vector retrieve by @chengmengli06 in #327

• [bugfix] fix sync train not exit bug due to uneven dataset shard caused by filter invalid samples by @chengmengli06 in #325

• [bugfix]: fix bug of embedding regularization by @yangxudong in #336

• [bug fix]: fix bug of multiple losses of rank model by @yangxudong in #335

• [bugfix]: fix the problem of can not display code snippets in document by @yangxudong in #341

• [bugfix]:fix layer for sequence feature by @lgqfhwy in #342

• [bugfix]: fix bug for graphlearn dist eval by @lgqfhwy in #334

• [bugfix]: fix a bug of loading checkpoint panic when graph meta is not written by @yangxudong in #346

• [bugfix]: fix ds osshdfsmnt output by @chenglongliu123 in #337

• [bugfix]: fix clear model, predict two model, sync on pai. by @lgqfhwy in #348

• [bugfix]: fix share input bug by @chengmengli06 in #313

• [bugfix]: fix dist eval ps high memory by @lgqfhwy in #318

• [bugfix]: fix dist eval ps high memory bug by @lgqfhwy in #302

• [bugfix]: fix negative sample using hive input bug by @chengmengli06 in #309

• [bugfix]: GAUC/SessionAuc when the id field is not configured in feature groups by @yangxudong in #310

• [bugfix]: improve ckpt restore performance by @chengmengli06 in #311

• [docs]: update docs by @chengmengli06 in #312

• [docs]: add docs for processor by @chengmengli06 in #314

• [docs] update fine_tune_checkpoint specification by @chengmengli06 in #317

New Contributors

• @wwxxzz made their first contribution in #361

Full Changelog: v0.5.6...v0.6.2

v0.6.0

What's Changed

• [bugfix]: update python publich ci yml by @chengmengli06 in #294
• [feat]: CSVPredictor support embedding input by @chenglongliu123 in #298
• [doc]: update nni hpo docs by @yjjinjie in #299
• [doc]: add docs for ev and sample_weight by @chengmengli06 in #305
• [bugfix]: Fix dist eval ps high memory bug by @lgqfhwy in #302
• update hpo docs and remove tensorflow in runtime.txt by @yjjinjie in #307
• [bugfix]: fix negative sample using hive input bug by @chengmengli06 in #309
• [bugfix]: of GAUC/Session AUC metric when the id field is not configurated in feature groups by @yangxudong in #315
• [bugfix]: improve ckpt restore performance by @chengmengli06 in #311
• [docs] update fine_tune_checkpoint specification by @chengmengli06 in #317
• [bugfix]: fix share input bug by @chengmengli06 in #313
• [feat]add udf for label by @dawn310826 in #303

Full Changelog: v0.5.6...v0.6.0

release/0.5.6

What's Changed

• [bugfix]: fix load_fg_json_to_config, make it reentrant by @chengmengli06 in #204

• [bugfix]: fix slow node crush in hive predictor by @dawn310826 in #220

• [bugfix]: fix bug of softmax_loss_with_negative_mining loss by @yangxudong in #237

• [bugfix]: fix bug for distribute eval worker hangs by @lgqfhwy in #252

• [bugfix]: fix embedding variable restore bug by @chengmengli06 in #234

• [bugfix]: fix share not used bug by @chengmengli06 in #233

• [bugfix]: fix bug of sequence feature rtp input by @yangxudong in #258

• [bugfix]: fix bug for sequence feature wide and deep by @lgqfhwy in #270

• [bugfix] export user_emb and item_emb used to calculate simularity by @cosmozhang1995 in #277

• [feature]: add docker file by @chengmengli06 in #211

• [feature]: support gzipped csv files by @chengmengli06 in #203

• [feature]: add hive rtp input by @dawn310826 in #200

• [feature]: add ngpu train scripts by @chengmengli06 in #213

• [feature]: add hadoop env in Dockerfile by @tiankongdeguiji in #214

• [feature]: update hive predictor: load predict result to hive by @dawn310826 in #215

• [feature]: add support for clear_model and clear_export option by @chengmengli06 in #218

• [feature]: add CMBF model by @yangxudong in #217

• [feature]: add nni hpo, json.load->yaml.safe.load,codestyle by @yjjinjie in #222

• [feature]: add CMBF layer for DBMTL model by @yangxudong in #223

• [feature]: support checkpoint_path dir by @dawn310826 in #225

• [feature]: optimize io cost of hiveinput and add hiveparquetinput by @dawn310826 in #224

• [feature]: modify document structure by @yangxudong in #226

• [feature]: add support for incremental_train.md by @chengmengli06 in #228

• [feature]: add support easyrec distribute metrics by @lgqfhwy in #201

• [feature]: add support for frequency filter and steps_to_alive for embedding variable by @chengmengli06 in #231

• [feature]: add incremental update for odl by @chengmengli06 in #257

• [feature]: support csv with header by @dawn310826 in #285

• [feature]: add designer doc by @chengmengli06 in #287

• [feature]: support gnn on datascience by @dawn310826 in #229

• [feature]: support hit rate on ds by @dawn310826 in #244

• [feature]: add user define eval_result_path by @lgqfhwy in #245

• [feature]: fix bug of fail to assign default value of rtp input field by @yangxudong in #232

• [feature]: add support for tf25 py3 test by @chengmengli06 in #255

• [feature]: support multi loss for multi task learning model by @yangxudong in #260

• [feature]: support only sequence feature && fix neg sampler bug for sequence feature by @lgqfhwy in #264

• [feature]: add aux_hist_seq by @lgqfhwy in #265

• [feature]: add Uniter model and Uniter layer for DBMTL model by @yangxudong in #266

• [feature]: add sequence feature negative sample process by @lgqfhwy in #267

• [feature]: features/cmbf by @yangxudong in #268

• [feature]: add dlc tutorial by @chengmengli06 in #272

• [feature]: fix rtp_native documentation: add reserve_default: false in fg.json by @cosmozhang1995 in #278

• [feature]: add pycharm vscode docker environment by @chengmengli06 in #293

New Contributors

• @yjjinjie made their first contribution in #222
• @chenglongliu123 made their first contribution in #279

Full Changelog: v0.4.7...v0.5.6