These commits are when the Protocol Buffers files have changed: (only the last 100 relevant commits are shown)
| Commit: | bd9aac8 | |
|---|---|---|
| Author: | Eric Ge | |
| Committer: | GitHub | |
dat model supports amm_loss_weight (#516)
The documentation is generated from this commit.
| Commit: | 72e5098 | |
|---|---|---|
| Author: | Eric Ge | |
| Committer: | GitHub | |
[feature] Dual Augmented Two-Towers (#503) * dat model implementation * DAT documentation * DAT unit test
| Commit: | f00d8a8 | |
|---|---|---|
| Author: | yangxudong | |
| Committer: | GitHub | |
add ZILN loss for ltv prediction task (#498) * add ZILN loss for ltv prediction task & add documents
| Commit: | 6686079 | |
|---|---|---|
| Author: | Eric Ge | |
| Committer: | GitHub | |
[feature]Add CIN module and Parallel DSSM model (#493) * match model supports backbone components * CIN module implementation * CIN module fix * CIN: add kernel/bias regularizer * add docs for parallel dssm model * add unit test for parallel dssm * fix CIN compatibility issue with tf1.x * fix tf.initializer issue * fix python2.7 string compatility issue
| Commit: | 7f006a9 | |
|---|---|---|
| Author: | Eric Ge | |
| Committer: | GitHub | |
Feature/backbone match model (#492) * match model supports backbone components * add DSSM/DSSM_SENet samples and unit tests * DSSM doc updates
| Commit: | 2df8f30 | |
|---|---|---|
| Author: | Eric Ge | |
| Committer: | GitHub | |
[feature] add a DSSM derivative model: integrate SENet into DSSM (#489) * DSSM + SENet * dssm_senet model implementation and documentation * update recall.rst * update senet docs * code style fix * code style fix * add dssm_senet unit test
| Commit: | f403b69 | |
|---|---|---|
| Author: | yangxudong | |
| Committer: | GitHub | |
add AITM component & upgrade NaryDisEmbedding & fix a few bugs (#487) * add AITM component & upgrade NaryDisEmbedding & fix a few bugs
| Commit: | 0c22985 | |
|---|---|---|
| Author: | 卫苏 | |
add AITM component & upgrade NaryDisEmbedding & fix a few bugs
| Commit: | f4948af | |
|---|---|---|
| Author: | yangxudong | |
| Committer: | GitHub | |
Add NaryDisEmbedding & Transformer Component (#485) * add transformer layer & NaryDisEmbedding
| Commit: | 491e7ea | |
|---|---|---|
| Author: | 卫苏 | |
add transformer layer
| Commit: | 1659947 | |
|---|---|---|
| Author: | 卫苏 | |
add transformer layer
| Commit: | dfe285e | |
|---|---|---|
| Author: | 卫苏 | |
add transformer layer
| Commit: | c648b38 | |
|---|---|---|
| Author: | 卫苏 | |
add narydis embedding
| Commit: | c1147b2 | |
|---|---|---|
| Author: | yangxudong | |
| Committer: | GitHub | |
Add support for componet based Custom op (#430) * add custom op & kd loss & add support for rank distillation
| Commit: | f015215 | |
|---|---|---|
| Author: | 卫苏 | |
Merge branch 'master' into custom_op
| Commit: | cbec539 | |
|---|---|---|
| Author: | Eric Ge | |
| Committer: | GitHub | |
Add temperature coefficient for DSSM model (#482) * add temperature coefficient for DSSM model * github actions_runner test * fix code style check * do normalization when use cosine loss * remove labelcomment in code style check * fix simi func in test case dssm_distribute_eval_reg
| Commit: | de96ccf | |
|---|---|---|
| Author: | 卫苏 | |
add support for rank distillation
| Commit: | 645b0dd | |
|---|---|---|
| Author: | weisu.yxd | |
add custom op & kd loss
| Commit: | 07b5ee3 | |
|---|---|---|
| Author: | weisu.yxd | |
add custom op & kd loss
| Commit: | 5e9f7a4 | |
|---|---|---|
| Author: | weisu.yxd | |
add custom op & kd loss
| Commit: | 1b1511b | |
|---|---|---|
| Author: | weisu.yxd | |
add custom op
| Commit: | f2f5b74 | |
|---|---|---|
| Author: | weisu.yxd | |
add custom op
| Commit: | ffda1da | |
|---|---|---|
| Author: | weisu.yxd | |
Merge branch 'master' into custom_op
| Commit: | 93886e5 | |
|---|---|---|
| Author: | yangxudong | |
| Committer: | GitHub | |
Add Highway & Attention Component And AITM model (#471) * modify highway layer & add attention layer and AITM model
| Commit: | 17c13e5 | |
|---|---|---|
| Author: | weisu.yxd | |
add attention layer and AITM model
| Commit: | 871a40e | |
|---|---|---|
| Author: | weisu.yxd | |
add attention layer and AITM model
| Commit: | 98b7c8c | |
|---|---|---|
| Author: | yangxudong | |
| Committer: | GitHub | |
fix bug of text cnn, add TextCNN component and model (#462) * fix bug of text cnn, add TextCNN component and model
| Commit: | def7901 | |
|---|---|---|
| Author: | chengmengli06 | |
| Committer: | GitHub | |
[feature]: add cpu kv support for sok (#455) * add support for cpu kv in sok * fix embedding parallel lookup batch_size bug * add conditional placement of feature column on cpu * add comments and update sok docker to support cpu kv
| Commit: | d504cf3 | |
|---|---|---|
| Author: | chengmengli06 | |
add comments and update docker
| Commit: | 93a25a6 | |
|---|---|---|
| Author: | chengmengli06 | |
remove test code
| Commit: | 89989c2 | |
|---|---|---|
| Author: | chengmengli06 | |
add embedding_on_cpu param
| Commit: | 2e0cd36 | |
|---|---|---|
| Author: | chengmengli06 | |
add support for cpu kv in sok
| Commit: | fdaa9a8 | |
|---|---|---|
| Author: | weisu.yxd | |
Merge branch 'master' of https://github.com/alibaba/EasyRec into custom_op Conflicts: docs/source/index.rst easy_rec/python/layers/backbone.py easy_rec/version.py
| Commit: | 1443930 | |
|---|---|---|
| Author: | chengmengli06 | |
| Committer: | GitHub | |
[feature]: add sok refactor (#446) * refactor sok dynamic variable; * add documents on embedding parallel; * add c++ load dense embed. Co-authored-by: 杨熙 <mengli.cml@alibaba-inc.com>
| Commit: | 4faeed6 | |
|---|---|---|
| Author: | chengmengli06 | |
add documents on embedding parallel
| Commit: | 315cc02 | |
|---|---|---|
| Author: | chengmengli06 | |
refactor sok dynamic variable
| Commit: | fddb73a | |
|---|---|---|
| Author: | chengmengli06 | |
| Committer: | GitHub | |
[feature]: add embedding parallel (#438) * add parquet input in packed format and non-packed format * add support for embedding parallel, shard embedding across workers, based on horovod all2all and allreduce * refactor predictor into separate files and add parquet predictors * add script for custom ops build * add adam sparse optimizer * add fast and memory efficient auc implementation
| Commit: | b58e246 | |
|---|---|---|
| Author: | chengmengli06 | |
limit parquet case to run on python3 and tf2.x
| Commit: | cc3f7d0 | |
|---|---|---|
| Author: | chengmengli06 | |
add comments on parquet input
| Commit: | 091476b | |
|---|---|---|
| Author: | chengmengli06 | |
fix merge conflicts
| Commit: | cb806b2 | |
|---|---|---|
| Author: | yangxudong | |
| Committer: | GitHub | |
Bug fix of SENet when run with MirroredStrategy (#443) * fix bug of SENet when run with tf.distribute.MirroredStrategy * add LayerNormalization Layer
| Commit: | b9d3a6f | |
|---|---|---|
| Author: | yzhjerry | |
add_bernoulli_mask
| Commit: | ba60b9c | |
|---|---|---|
| Author: | tiankongdeguiji | |
| Committer: | GitHub | |
[feat]: add optional safe embedding & fix logging problem (#437)
| Commit: | aab9cab | |
|---|---|---|
| Author: | chengmengli06 | |
fix unit test bug
| Commit: | c349303 | |
|---|---|---|
| Author: | chengmengli06 | |
remove debug code
| Commit: | 23a379f | |
|---|---|---|
| Author: | chengmengli06 | |
merge master and remove extra files
| Commit: | c7f9100 | |
|---|---|---|
| Author: | 天邑 | |
add auto_multi_value
| Commit: | cf1159b | |
|---|---|---|
| Author: | chengmengli06 | |
merge master and remove extra files
| Commit: | 54580b1 | |
|---|---|---|
| Author: | chengmengli06 | |
add support not pack case
| Commit: | 75a97ec | |
|---|---|---|
| Author: | tiankongdeguiji | |
| Committer: | GitHub | |
[feat]: add parquet dataset (#435) add parquet_input_v3
| Commit: | 9028389 | |
|---|---|---|
| Author: | 天邑 | |
| Committer: | 天邑 | |
add parquet_input_v3
| Commit: | ca35cba | |
|---|---|---|
| Author: | 天邑 | |
add parquet_input_v3
| Commit: | 1fb889d | |
|---|---|---|
| Author: | yangxudong | |
| Committer: | GitHub | |
[feat]: add ppnet & upgrade backbone (#434) * add ppnet & upgrade backbone, mmoe&mask block component
| Commit: | 3ac7494 | |
|---|---|---|
| Author: | chengmengli06 | |
support old training path
| Commit: | e3b91a5 | |
|---|---|---|
| Author: | chengmengli06 | |
add support for ALL_COLUMNS in parquet predictor
| Commit: | b5a143e | |
|---|---|---|
| Author: | chengmengli06 | |
| Committer: | chengmengli06 | |
fix bug
| Commit: | 5477cb2 | |
|---|---|---|
| Author: | chengmengli06 | |
add adam sparse
| Commit: | bc9f2f5 | |
|---|---|---|
| Author: | weisu.yxd | |
add custom op demo
| Commit: | bbfe530 | |
|---|---|---|
| Author: | chengmengli06 | |
add drop_remainder option as required by sok
| Commit: | ae74f18 | |
|---|---|---|
| Author: | chengmengli06 | |
add sok debug case2
| Commit: | bd5bd02 | |
|---|---|---|
| Author: | chengmengli06 | |
add adam_s
| Commit: | 54c9e9e | |
|---|---|---|
| Author: | 杨熙 | |
add support for keep model
| Commit: | e0a8a1d | |
|---|---|---|
| Author: | weisu.yxd | |
add custom op demo
| Commit: | e18caf0 | |
|---|---|---|
| Author: | yangxudong | |
| Committer: | GitHub | |
[bugfix]: fix a bug of dice activation of MLP layer (#428) * fix bug of Dice activation of MLP layer
| Commit: | 2ae081d | |
|---|---|---|
| Author: | chengmengli06 | |
| Committer: | GitHub | |
add support for separate optimizers for the embedding layer (#424)
| Commit: | d2f81ae | |
|---|---|---|
| Author: | chengmengli06 | |
add support for separate optimizers for the embedding layer
| Commit: | 165838e | |
|---|---|---|
| Author: | yangxudong | |
| Committer: | GitHub | |
Add support for contrastive learning & add CL4SRec model (#421) * add support for contrastive learning * add contrastive learning CL4SRec --------- Co-authored-by: yzhjerry <978409300@qq.com>
| Commit: | f8ccdd9 | |
|---|---|---|
| Author: | weisu | |
add contrastive learning CL4SRec
| Commit: | 1e065b1 | |
|---|---|---|
| Author: | weisu | |
add support for contrastive learning
| Commit: | 73d3972 | |
|---|---|---|
| Author: | yangxudong | |
| Committer: | GitHub | |
Bug fix of `package` syntax of backbone network，support layer reuse of components (#419) * fix bug of `package` syntax of backbone * add document and test cases of BST & DIN backbone model * add powerful support for contrastive learning
| Commit: | eea8a8b | |
|---|---|---|
| Author: | yzhjerry | |
add cl
| Commit: | 0db518a | |
|---|---|---|
| Author: | yzhjerry | |
contrast learning
| Commit: | 7c94b60 | |
|---|---|---|
| Author: | wwxxzz | |
| Committer: | GitHub | |
Add pdn model (#409) * add PDN model,config,data * add PDN to intro.md * modify r.t. reviews * modify images of pdn * modify md of pdn
| Commit: | a59ee14 | |
|---|---|---|
| Author: | chengmengli06 | |
add support for non sharing of embedding between groups
| Commit: | 5c93ed7 | |
|---|---|---|
| Author: | wwxxzz | |
add PDN model,config,data
| Commit: | 1be51db | |
|---|---|---|
| Author: | yangxudong | |
| Committer: | GitHub | |
[feature]: Add component-based develop Backbone Network for EasyRec (#388) * [feat] add component-based development for easy_rec
| Commit: | ad76a86 | |
|---|---|---|
| Author: | weisu | |
[feat] modify wide_output_dim
| Commit: | 6ac8a0c | |
|---|---|---|
| Author: | chengmengli06 | |
add support for sok train in sync mode
| Commit: | f87395b | |
|---|---|---|
| Author: | weisu | |
Merge branch 'master' of https://github.com/alibaba/EasyRec into backbone Conflicts: easy_rec/python/input/input.py easy_rec/python/layers/input_layer.py easy_rec/version.py
| Commit: | faf8ddf | |
|---|---|---|
| Author: | weisu | |
upgrade to new version
| Commit: | 303eef3 | |
|---|---|---|
| Author: | chengmengli06 | |
| Committer: | GitHub | |
[feature] add support for multival separation using combo feature (#396) * add support for multival separation using combo feature * add support for git lfs auto pull * make larger the duration used in the save checkpoints seconds test, which is easy to fail under tf 1.12
| Commit: | 7648671 | |
|---|---|---|
| Author: | chengmengli06 | |
| Committer: | GitHub | |
[feature] add support for horovod (#389) * add support for horovod * remove dependency on HOROVOD_RANK env
| Commit: | 6ed390c | |
|---|---|---|
| Author: | chengmengli06 | |
update combo input sep and join sep
| Commit: | 5a4a1b7 | |
|---|---|---|
| Author: | chengmengli06 | |
add support for multival separation using combo feature
| Commit: | 6ea2bcf | |
|---|---|---|
| Author: | weisu | |
[feat] enhance backbone
| Commit: | 2c591fc | |
|---|---|---|
| Author: | weisu | |
fix a bug
| Commit: | e9311c5 | |
|---|---|---|
| Author: | weisu | |
[feat] add component-base dev support for multi task model
| Commit: | 58a6bf9 | |
|---|---|---|
| Author: | weisu | |
[feat] add block package for building a reusable DAG
| Commit: | 2847347 | |
|---|---|---|
| Author: | weisu | |
add block package for reuse sub network
| Commit: | b60ed2c | |
|---|---|---|
| Author: | weisu | |
[feat] modify input format
| Commit: | 1e785b4 | |
|---|---|---|
| Author: | chengmengli06 | |
remove dependency on HOROVOD_RANK env
| Commit: | a9aff75 | |
|---|---|---|
| Author: | weisu | |
add gate layer
| Commit: | 7ddaa90 | |
|---|---|---|
| Author: | chengmengli06 | |
bugfix
| Commit: | 4179b27 | |
|---|---|---|
| Author: | chengmengli06 | |
add support for separate dnn
| Commit: | 851c48b | |
|---|---|---|
| Author: | weisu | |
[feat] add component-based development for easy_rec
| Commit: | 0c85dd2 | |
|---|---|---|
| Author: | weisu | |
add gate layer
| Commit: | ac56512 | |
|---|---|---|
| Author: | chengmengli06 | |
add support for retarget
| Commit: | 31bbe04 | |
|---|---|---|
| Author: | chengmengli06 | |
add support for sparse attention
| Commit: | c4f5ea9 | |
|---|---|---|
| Author: | weisu | |
[feat]: format backbone code, add recurrent and sequential layer
| Commit: | 7d0e350 | |
|---|---|---|
| Author: | weisu | |
[feat]: add more backbone blocks