Introduction
DeepRec is a high-performance recommendation deep learning framework based on TensorFlow 1.15, Intel-TensorFlow and NVIDIA-TensorFlow. It is hosted in incubation in LF AI & Data Foundation.
Background
Recommendation models have huge commercial values for areas such as retailing, media, advertisements, social networks and search engines. Unlike other kinds of models, recommendation models have large amount of non-numeric features such as id, tag, text and so on which lead to huge parameters.
DeepRec has been developed since 2016, which supports core businesses such as Taobao Search, recommendation and advertising. It precipitates a list of features on basic frameworks and has excellent performance in recommendation models training and inference. So far, in addition to Alibaba Group, dozens of companies have used DeepRec in their business scenarios.
Key Features
DeepRec has super large-scale distributed training capability, supporting recommendation model training of trillion samples and over ten trillion parameters. For recommendation models, in-depth performance optimization has been conducted across CPU and GPU platform. It contains list of features to improve usability and performance for super-scale scenarios.
Embedding & Optimizer
- Embedding Variable.
- Dynamic Dimension Embedding Variable.
- Adaptive Embedding Variable.
- Multiple Hash Embedding Variable.
- Multi-tier Hybrid Embedding Storage.
- Group Embedding.
- AdamAsync Optimizer.
- AdagradDecay Optimizer.
Training
- Asynchronous Distributed Training Framework (Parameter Server), such as grpc+seastar, FuseRecv, StarServer etc.
- Synchronous Distributed Training Framework (Collective), such as HybridBackend, Sparse Operation Kits (SOK) etc.
- Runtime Optimization, such as Graph Aware Memory Allocator (GAMMA), Critical-path based Executor etc.
- Runtime Optimization (GPU), GPU Multi-Stream Engine which support multiple CUDA compute stream and CUDA Graph.
- Operator level optimization, such as BF16 mixed precision optimization, embedding operator optimization and EmbeddingVariable on PMEM and GPU, new hardware feature enabling, etc.
- Graph level optimization, such as AutoGraphFusion, SmartStage, AutoPipeline, Graph Template Engine, Sample-awared Graph Compression, MicroBatch etc.
- Compilation optimization, support BladeDISC, XLA etc.
Deploy and Serving
- Delta checkpoint loading and exporting.
- Super-scale recommendation model distributed serving.
- Multi-tier hybrid storage and multi backend supported.
- Online deep learning with low latency.
- High performance inference framework SessionGroup (share-nothing), with multiple threadpool and multiple CUDA stream supported.
- Model Quantization.
Installation
Prepare for installation
CPU Platform
alideeprec/deeprec-build:deeprec-dev-cpu-py38-ubuntu20.04
GPU Platform
alideeprec/deeprec-build:deeprec-dev-gpu-py38-cu116-ubuntu20.04
How to Build
Configure
$ ./configure
Compile for CPU and GPU defaultly
$ bazel build -c opt --config=opt //tensorflow/tools/pip_package:build_pip_package
Compile for CPU and GPU: ABI=0
$ bazel build --cxxopt="-D_GLIBCXX_USE_CXX11_ABI=0" --host_cxxopt="-D_GLIBCXX_USE_CXX11_ABI=0" -c opt --config=opt //tensorflow/tools/pip_package:build_pip_package
Compile for CPU optimization: oneDNN + Unified Eigen Thread pool
$ bazel build -c opt --config=opt --config=mkl_threadpool //tensorflow/tools/pip_package:build_pip_package
Compile for CPU optimization and ABI=0
$ bazel build --cxxopt="-D_GLIBCXX_USE_CXX11_ABI=0" --host_cxxopt="-D_GLIBCXX_USE_CXX11_ABI=0" -c opt --config=opt --config=mkl_threadpool //tensorflow/tools/pip_package:build_pip_package
Create whl package
$ ./bazel-bin/tensorflow/tools/pip_package/build_pip_package /tmp/tensorflow_pkg
Install whl package
$ pip3 install /tmp/tensorflow_pkg/tensorflow-1.15.5+${version}-cp38-cp38m-linux_x86_64.whl
Latest Release Images
Image for CPU
alideeprec/deeprec-release:deeprec2304-cpu-py38-ubuntu20.04
Image for GPU CUDA11.6
alideeprec/deeprec-release:deeprec2304-gpu-py38-cu116-ubuntu20.04
Continuous Build Status
Official Build
| Build Type | Status |
|---|---|
| Linux CPU | |
| Linux GPU | |
| Linux CPU Serving | |
| Linux GPU Serving |
Official Unit Tests
User Document
Chinese: https://deeprec.readthedocs.io/zh/latest/
English: https://deeprec.readthedocs.io/en/latest/
Contact Us
Join the Official Discussion Group on DingTalk
Join the Official Discussion Group on WeChat