Context-aware Fast Food Recommendation with Ray on Apache Spark at Burger King
0 likes327 views
This document presents a context-aware fast food recommendation system implemented at Burger King using Ray on Apache Spark. It outlines the challenges in food recommendation, the deployment of a session-based recommendation model, and performance metrics indicating an improvement in conversion rates. Additionally, it discusses project ORCA, which facilitates scalable distributed training and inference across large clusters.
![Minimum code changes and learning efforts are needed to scale the training from single node
to big data clusters.
The entire pipeline runs on a single cluster. No extra data transfer needed.
Project Orca provides a user-friendly interface for the pipeline.
End-to-end Distributed Training Pipeline
from zoo.orca import init_orca_context
from zoo.orca.learn.mxnet import Estimator
# init_orca_context unifies SparkContext and RayContext
sc = init_orca_context(cluster_mode="yarn", num_nodes, cores, memory)
# Use sc to load data and do data preprocessing.
mxnet_estimator = Estimator(train_config, model=txt, loss=SoftmaxCrossEntropyLoss(),
metrics=[mx.metric.Accuracy(), mx.metric.TopKAccuracy(3)])
mxnet_estimator.fit(data=train_rdd, validation_data=val_rdd, epochs=…, batch_size=…)](https://image.slidesharecdn.com/213huangwang-201130191526/85/Context-aware-Fast-Food-Recommendation-with-Ray-on-Apache-Spark-at-Burger-King-19-320.jpg)
Context-aware Fast Food Recommendation with Ray on Apache Spark at Burger King
- 1. Context-aware Fast Food Recommendation with Ray on Apache Spark at Burger King LUYANG WANG Burger King Corporation KAI HUANG Intel Corporation
- 2. Agenda LUYANG WANG Food recommendation use case TxT model in detail KAI HUANG AI on big data Distributed training pipeline with Ray on Apache Spark
- 3. Food Recommendation Use Case
- 4. Food Recommendation Use Case Guest arrives ODMB Checks Menu Board Cashier enters order Checks Menu Board Guest completes order
- 5. Food Recommendation Use Case Guest arrives ODMB Checks Menu Board Cashier enters order Checks Menu Board Guest completes order
- 6. Use Case Challenges Lack of user identifiers Same session food compatibilities Other variables in our use case: locations, weathers, time, etc. Deployment challenges Challenges
- 7. Use Case Challenges Session based recommendation model Able to take complex context features into consideration Able to be deployed anywhere, both edge / cloud Solutions
- 8. Transformer Cross Transformer (TxT)
- 9. TxT Model Overview Sequence Transformer ▪ Taking item order sequence as input Context Transformer ▪ Taking multiple context features as input Latent Cross Joint Training ▪ Element-wise product for both transformer outputs Model Components
- 10. Model Comparison RNN Latent CrossTxT
- 11. Offline Evaluation Offline Training ResultOffline Training Loss Model Top1 Accuracy Top3 Accuracy RNN 29.98% 46.24% Contextual ItemCF 32.18% 48.37% RNN Latent Cross 33.10% 49.98% TxT 34.52% 52.37%
- 12. Online Performance A/B Testing ResultInference Performance Model Conversation Rate Gain Add-on Sales Gain RNN Latent Cross (control) - - TxT +7.5% +4.7% 18 20 0 5 10 15 20 25 RNN Latent Cross TxT Inference Latency (ms) Inference Latency (ms)
- 14. AI on Big Data
- 15. AI on Big Data BigDL: Distributed Deep Learning Framework for Apache Spark https://github.com/intel-analytics/BigDL Analytics Zoo: Distributed TensorFlow, Keras and PyTorch on Apache Spark/Flink & Ray https://github.com/intel-analytics/analytics-zoo We develop Project Orca in Analytics Zoo based on Spark and Ray to allow users to easily scale out single node Python notebook across large clusters, by providing: ▪ Data-parallel preprocessing for Python AI (supporting common Python libraries such as Pandas, Numpy, PIL, TensorFlow Dataset, PyTorch DataLoader, etc.) ▪ Sklearn-style APIs for transparently distributed training and inference (supporting TensorFlow, PyTorch, Keras, MXNet, Horovod, etc.) https://github.com/intel-analytics/analytics-zoo/tree/master/pyzoo/zoo/orca Accelerating Data Analytics + AI Solutions At Scale
- 16. Ray Tune: Scalable Experiment Execution and Hyperparameter Tuning RLlib: Scalable Reinforcement Learning RaySGD: Distributed Training Wrappers https://github.com/ray-project/ray/ Ray is a fast and simple framework for building and running distributed applications. Ray is packaged with several high-level libraries to accelerate machine learning workloads. Ray Core provides easy Python interface for parallelism by using remote functions and actors.
- 17. Distributed Training Pipeline on Big Data
- 18. Runtime cluster environment preparation. Create a SparkContext on the drive node and use Spark to perform data cleaning, ETL, and preprocessing tasks. RayContext on Spark driver launches Ray across the cluster. Similar to RaySGD, we implement a lightweight shim layer around native MXNet modules for easy deployment on YARN cluster. Each MXNet worker takes the local data partition of Spark RDD or DataFrame from the plasma object store used by Ray. Seamlessly integrate Ray applications into Spark data processing pipelines. RayOnSpark
- 19. Minimum code changes and learning efforts are needed to scale the training from single node to big data clusters. The entire pipeline runs on a single cluster. No extra data transfer needed. Project Orca provides a user-friendly interface for the pipeline. End-to-end Distributed Training Pipeline from zoo.orca import init_orca_context from zoo.orca.learn.mxnet import Estimator # init_orca_context unifies SparkContext and RayContext sc = init_orca_context(cluster_mode="yarn", num_nodes, cores, memory) # Use sc to load data and do data preprocessing. mxnet_estimator = Estimator(train_config, model=txt, loss=SoftmaxCrossEntropyLoss(), metrics=[mx.metric.Accuracy(), mx.metric.TopKAccuracy(3)]) mxnet_estimator.fit(data=train_rdd, validation_data=val_rdd, epochs=…, batch_size=…)
- 20. Conclusion Context-Aware Fast Food Recommendation at Burger King with RayOnSpark https://arxiv.org/abs/2010.06197 https://medium.com/riselab/context-aware-fast-food-recommendation-at-burger-king- with-rayonspark-2e7a6009dd2d For more details of RayOnSpark: https://databricks.com/session_na20/running- emerging-ai-applications-on-big-data-platforms-with-ray-on-apache-spark More information for Analytics Zoo at: https://github.com/intel-analytics/analytics-zoo https://analytics-zoo.github.io/
- 21. Feedback Your feedback is important to us. Don’t forget to rate and review the sessions.