Download as PDF, PPTX
Uploaded byMLconf
Rajat Monga, Engineering Director, TensorFlow, Google at MLconf 2016
This document provides an overview of TensorFlow, an open source machine learning framework. It discusses how machine learning systems can become complex with modeling complexity, heterogeneous systems, and distributed systems. It then summarizes key aspects of TensorFlow, including its architecture, platforms, languages, parallelism approaches, algorithms, and tooling. The document emphasizes that TensorFlow handles complexity so users can focus on their machine learning ideas.
More Related Content
Similar to Rajat Monga, Engineering Director, TensorFlow, Google at MLconf 2016
![Coded Agents – with UiPath SDK + LangGraph [Virtual Hands-on Workshop]](https://cdn.slidesharecdn.com/ss_thumbnails/codedagentsdeck-251215155422-5497c599-thumbnail.jpg?width=640&height=640&fit=bounds)
Rajat Monga, Engineering Director, TensorFlow, Google at MLconf 2016
- 1. Deep Learning withTensorFlow Rajat Monga Engineering Director, TensorFlow
- 3. How complex are MachineLearning Systems?
- 4.
- 5.
- 6.
- 7.
- 8. That’s not all... Input batch ... Parameters Labels PredictionLoss Update Parameter Server
- 9. That’s not all... Input batch ... ParametersUpdate Model Worker
- 10. UpdateUpdateUpdate Model That’s not all... Input batch ... ParametersUpdate Model Model Model ... Parallelize this! Distribute this!
- 11. Raspberry Pi Android iOS Portability isa requirement TPU GPU CPU
- 12. machine learning getscomplex quickly Modeling complexityHeterogenous System Distributed System
- 14. repository for “machine learning” categoryon GitHub #1 Released in Nov. 2015
- 15. 10,000+ commits sinceNov, 2015 490+ contributors 1M+ binary downloads #20 most popular repository on GitHub by stars Used in ML classes at many universities: Toronto, Berkeley, Stanford, ... Some Stats
- 18. TensorFlow powered CucumberSorter From: http://workpiles.com/2016/02/tensorflow-cnn-cucumber/
- 19. machine learning getscomplex quickly Modeling complexityHeterogenous System Distributed System
- 20.
- 21.
- 22. TensorFlow Architecture CPU GPUAndroid iOS ... C++ front end Python front end ... Kernels Bindings + Compound Ops add mul Print reshape ... Core TensorFlow Execution System Ops
- 23.
- 24. machine learning getscomplex quickly Modeling complexityHeterogenous System Distributed System
- 25. Parallelism in Opimplementations MatMul ... ...
- 26. ... ... ... ... ...MatMul MatMul Task Parallelism inDataFlow graph
- 27.
- 28.
- 29. GPU 0 CPU AddMul biases learning rate −=... ... Distribution across Devices
- 30. GPU 0 CPU AddMul biases learning rate −=... ... Distribution ● TensorFlow inserts Send/Recv Ops to transport tensors across devices ● Recv ops pull data from Send ops Send Recv
- 31. GPU 0 CPU AddMul biases learning rate −=... ... Distribution ● TensorFlow inserts Send/Recv Ops to transport tensors across devices ● Recv ops pull data from Send ops Send Recv Send Recv Send Recv Send Recv
- 32.
- 33. machine learning getscomplex quickly Modeling complexityHeterogenous System Distributed System
- 34. MatMul Add Relu biases weights input input =... biases = tf.get_variable(’biases’, ...) weights = tf.get_variable(’weights’, ...) out = tf.matmul(input, weights) out = tf.add(out, biases) out = tf.nn.relu(out)
- 35. MatMul Add Relu biases weights input input =... output = tf.contrib.layers.fully_connected(input, ...)
- 36. Align cognitive modelwith programming model Each box is one line of code!
- 37. TensorFlow contains completealgorithms Linear{Classifier,Regressor} DNN{Classifier,Regressor} DNNLinearCombined{Classifier,Regressor} SVM KMeansClustering GMM ...
- 38. Simple machine learning classifier= learn.LinearClassifier(feature_columns=feature_columns, n_classes=10) classifier.fit(data, labels, batch_size=100, steps=1000) classifier.evaluate(eval_data, eval_labels) Tooling provided for distributed training and evaluation, graphical debugging, and export to production server (tensorflow/serving).
- 39.
- 40. TensorFlow handles complexityfor you... ...so you can focus on your ideas
- 41. 41 Ready to useMachine Learning models Use your own data to train models Cloud Vision API Cloud Translate API Cloud Machine Learning Develop - Model - Test Google BigQuery Cloud Storage Cloud Datalab Beta GA BetaGA Beta GA GA Cloud Natural Language API Beta Cloud Speech API TensorFlow
- 42.
- 44.
- 45.
- 46.