Enable DeepRec Features in the code and the features are shown below. The features have been enabled in WDL(https://github.com/changqi1/DeepRec/pull/37), and please notice that the comments can be mapped to the features below. Add the flags to enable/disable the features in the code.
If there is any problem when enabling the feature below, please describe the details of how to reproduce and what is the issue, especially the known issues below we have submitted to Alibaba.
Other Features : Disabled by default and test pass is fine. Don't need to ensure the same performance as before. This feature is not supported in feature_column API. We are waiting for Alibaba's update.
6) Multi-Hash Variable
Test
All of the features needs to be enabled in the code by adding flags.(WDL is the template)
Feature8~15 needs to be enabled by default and test passed with the same performance as before.
Other Features need to pass test, not ensure performance. Some of the features have known issues we submitted. If not passed, describe it clearly.
Other Requirements: Dockerfile and Documents
Waiting for Alibaba's requirements
Code Style and commit
Python: Keep aligned with DeepRec code.
Maintain
All of the issues and bugs related to this model need to be covered in the future.
Definition of Done
Run successfully in DeepRec and could get the same performance as the code before rebuilding.
Integrated into DeepRec successfully and commit the code follow DeepRec commit standard.
Rebuild MMoE to update API and Enable DeepRec Features Goal Rebuild MMoE to update API and enable DeepRec Features.
Requirement Details
Features list Enable the following DeepRec feature(Docs about the features from Alibaba https://deeprec.readthedocs.io/zh/latest/index.html):
Enabled By Default and test the AUC/ACC/Gsteps, which needs to be close to the result before rebuilding
_8) Auto Micro Batch same with https://github.com/alibaba/DeepRec/issues/127 9) FusedEmbedding API, embedding fusion 10) Smart Stage same with https://github.com/alibaba/DeepRec/issues/122 11) Auto Graph Fusion https://github.com/alibaba/DeepRec/issues/144 12) CPU Memory Optimization:START_STATISTIC_STEP, STOP_STATISTICSTEP, jemalloc 14) AdamAsync Optimizer 15) BF16
Disabled by default and test pass is fine. Don't need to ensure the same performance as before
1) Embedding Variable 7) GRPC++ and StarServer 13) Incremental Checkpoint 14) AdagradDecay 2) EmbeddingVariable advanced features：Embedding Elimination 3) EmbeddingVariable advanced feature：Embedding Filter 4) Dynamic-dimension Embedding Variable 5) Adaptive Embedding 17) WorkQueue
6) Multi-Hash Variable
Test
Other Requirements: Dockerfile and Documents
Code Style and commit
Maintain
Definition of Done