DeepCTR Torch Versions Save

【PyTorch】Easy-to-use,Modular and Extendible package of deep-learning based CTR models.

v0.2.9

1 year ago

Change Log

Major Features and Improvements

  1. Add multi-task models: SharedBottom, ESMM, MMOE, PLE @zanshuxun
  2. Support Python 3.9 and 3.10 @shenweichen

Bug Fixes and Other Changes

  • fix #232 #240 @zanshuxun

DisscussionGroup

公众号:浅梦学习笔记 微信:deepctrbot 学习小组 加入 主题集合
公众号 微信 学习小组

v0.2.8

1 year ago

Change Log

Bug Fixes and Other Changes

  • Fix model save error when using acc metric #106 @gaohongkui
  • Fix model error when all feature columns are dense #192 @zanshuxun

DisscussionGroup

公众号:浅梦学习笔记 微信:deepctrbot 学习小组 加入 主题集合
公众号 微信 学习小组

v0.2.7

2 years ago

Change Log

Major Features and Improvements

  • Add AFN model @WeiyuCheng

Bug Fixes and Other Changes

  • Fix callback error #171 @zanshuxun
  • Fix VarLenSparseFeat error #176 #179 #180 @zanshuxun
  • Change the output shape of InteractingLayer in Autoint #74 @zanshuxun

v0.2.6

3 years ago

Change Log

Major Features and Improvements

v0.2.5

3 years ago

Change Log

Bug Fixes and Other Changes

  • fix error in dcn-mix

v0.2.4

3 years ago

Change Log

Major Features and Improvements

Bug Fixes and Other Changes

  • Solve issues in #69, #87, #93, #128, #130
  • Fix regularization error in pytorch1.7,
  • improve compatibility for old version

API changes

  • Add History in deepctr_torch.callbacks

v0.2.3

3 years ago

Change Log

Major Features and Improvements

Bug Fixes and Other Changes

  • Use float64 in metric to prevent nan/inf loss when calculating logloss

API changes

  • DCN add a new parameter:
    cross_parameterization: string, "vector" or "matrix", way to parameterize the cross network.if set to "matrix" then it will be DCN-M
  • Add EarlyStopping and ModelCheckpoint in deepctr_torch.callbacks

v0.2.2

3 years ago

Change Log

Major Features and Improvements

  • Improve the reproducibility of models

Bug Fixes and Other Changes

Fix RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.cuda.FloatTensor [128]], which is output 0 of SelectBackward, is at version 2; expected version 1 instead. Hint: enable anomaly detection to find the operation that failed to compute its gradient, with torch.autograd.set_detect_anomaly(True) in : https://github.com/shenweichen/DeepCTR-Torch/issues/88 https://github.com/shenweichen/DeepCTR-Torch/issues/98 https://github.com/shenweichen/DeepCTR-Torch/issues/90 https://github.com/shenweichen/DeepCTR-Torch/issues/102

v0.2.1

4 years ago

Change Log

Major Features and Improvements

Add DIN and DIEN

Bug Fixes and Other Changes

  • fix bug of #63 softmax dim error in InteractingLayer

v0.2.0

4 years ago

Change Log

Major Features and Improvements

  • Refactor feature columns.

  • Different features can use different embedding_dim

  • Add linear part to some models

  • Add SequencePoolingLayer(API)

Bug Fixes and Other Changes

  • Support double precision in metric calculation by setting use_double=True in model.fit() #15

API changes

  • embedding_size parameter of models is removed.Now we must set embedding_dim(default 4) in SparseFeat or VarLenSparseFeat .