https://github.com/alibaba/torcheasyrec
An easy-to-use framework for large scale recommendation algorithms.
https://github.com/alibaba/torcheasyrec
deepfm din dlrm dssm mind multi-task-learning recommendation-algorithms recommender-system tdm
Last synced: 6 months ago
JSON representation
An easy-to-use framework for large scale recommendation algorithms.
- Host: GitHub
- URL: https://github.com/alibaba/torcheasyrec
- Owner: alibaba
- License: apache-2.0
- Created: 2024-09-25T01:16:02.000Z (over 1 year ago)
- Default Branch: master
- Last Pushed: 2025-09-24T03:00:33.000Z (6 months ago)
- Last Synced: 2025-09-24T05:21:14.050Z (6 months ago)
- Topics: deepfm, din, dlrm, dssm, mind, multi-task-learning, recommendation-algorithms, recommender-system, tdm
- Language: Python
- Homepage: https://torcheasyrec.readthedocs.io
- Size: 10.6 MB
- Stars: 226
- Watchers: 7
- Forks: 41
- Open Issues: 4
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# TorchEasyRec Introduction
## What is TorchEasyRec?
### TorchEasyRec is an easy-to-use framework for Recommendation
TorchEasyRec implements state of the art deep learning models used in common recommendation tasks: candidate generation(matching), scoring(ranking), multi-task learning and generative recommendation. It improves the efficiency of generating high performance models by simple configuration and easy customization.
### Get Started
- [Local](docs/source/quick_start/local_tutorial.md)
- [PAI-DLC](docs/source/quick_start/dlc_tutorial.md)
## Why TorchEasyRec?
### Run everywhere
- Local / [PAI-DLC](https://help.aliyun.com/zh/pai/user-guide/container-training) / [PAI-DSW](https://help.aliyun.com/zh/pai/user-guide/dsw-notebook-service) / [EMR-DataScience](https://help.aliyun.com/document_detail/170836.html)
### Diversified input data
- [MaxCompute Table](https://help.aliyun.com/document_detail/27819.html)
- [OSS files](https://help.aliyun.com/product/31815.html)
- CSV files
- Parquet files
### Easy-to-use
- Flexible feature config and model config
- Easy to implement [customized models](docs/source/models/user_define.md)
- Easy deployment to [EAS](https://help.aliyun.com/zh/pai/user-guide/eas-model-serving): automatic scaling, easy monitoring
### Fast and robust
- Efficient and robust feature generation
- Large scale embedding with different sharding strategies
- Hybrid data-parallelism/model-parallelism
- Optimized kernels for RecSys powered by TorchRec
- Mixed precision
- Consistency guarantee: train and serving
### A variety of features & models
- IdFeature / RawFeature / ComboFeature / LookupFeature / MatchFeature / ExprFeature / KvDotProduct / BoolMaskFeature / OverlapFeature / TokenizeFeature / SequenceIdFeature / SequenceRawFeature / SequenceFeature
- Match: [DSSM](docs/source/models/dssm.md) / [TDM](docs/source/models/tdm.md) / [DAT](docs/source/models/dat.md) / [MIND](docs/source/models/mind.md)
- Rank: [WideAndDeep](docs/source/models/wide_and_deep.md) / [DeepFM](docs/source/models/deepfm.md) / [MultiTower](docs/source/models/multi_tower.md) / [DIN](docs/source/models/din.md) / [RocketLaunching](docs/source/models/rocket_launching.md) / [DLRM](docs/source/models/dlrm.md) / [MaskNet](docs/source/models/masknet.md) / [DCN](docs/source/models/dcn.md) / [DCNv2](docs/source/models/dcn_v2.md) / [xDeepFM](docs/source/models/xdeepfm.md)
- Multi-Task: [MMoE](docs/source/models/mmoe.md) / [DBMTL](docs/source/models/dbmtl.md) / [PLE](docs/source/models/ple.md)
- Generative-Rec: [DlrmHSTU](docs/source/models/dlrm_hstu.md)
- More models in development
## Contribute
Any contributions you make are greatly appreciated!
- Please report bugs by submitting a issue.
- Please submit contributions using pull requests.
- Please refer to the [Development](docs/source/develop.md) document for more details.
## Contact
### Join Us
- DingDing Group: 32260796, click [this url](https://page.dingtalk.com/wow/z/dingtalk/simple/ddhomedownload?action=joingroup&code=v1,k1,MwaiOIY1Tb2W+onmBBumO7sQsdDOYjBmv6FXC6wTGns=&_dt_no_comment=1&origin=11#/) or scan QrCode to join!
- DingDing Group2: 37930014162, click [this url](https://page.dingtalk.com/wow/z/dingtalk/simple/ddhomedownload?action=joingroup&code=v1,k1,1ppFWEXXNPyxUClHh77gCmpfB+JcPhbFv6FXC6wTGns=&_dt_no_comment=1&origin=11#/) or scan QrCode to join!
- Email Group: easy_rec@service.aliyun.com.
### Enterprise Service
- If you have any questions about how to use TorchEasyRec, please join the DingTalk group and contact us.
- If you have enterprise service needs or need to purchase Alibaba Cloud services to build a recommendation system, please join the DingTalk group to contact us.
## License
TorchEasyRec is released under Apache License 2.0. Please note that third-party libraries may not have the same license as TorchEasyRec.