Past Meetup

Deep Dive into Google TPU, TFRecord, Dataset API, Kafka, Math Behind Neural Nets

This Meetup is past

630 people went

Details

PLEASE RSVP HERE TO MAKE SURE YOU GET A SPOT:

https://www.eventbrite.com/e/deep-dive-into-google-tpu-tfrecord-dataset-api-kafka-math-behind-neural-nets-tickets-44304797843

THIS EVENT WILL BE RECORDED AND POSTED TO THE FOLLOWING:

* https://pipeline.ai
* https://youtube.pipeline.ai
* https://slideshare.pipeline.ai
* https://quickstart.pipeline.ai
* http://community.pipeline.ai

Agenda

Talk 0: Meetup Updates and Announcements (by Chris Fregly, Founder @ PipelineAI)

Talk 1: Comparing Spark and TensorFlow Model Training and Serving Pipelines including the Estimator API, Dataset API, TFRecord File Format, GPUs, and TPUs (by Chris Fregly, Founder @ PipelineAI)

https://www.linkedin.com/in/cfregly/

Related Links
* https://docs.google.com/presentation/d/16kHNtQslt-yuJ3w8GIx-eEH6t_AvFeQOchqGRFpAD7U/

* https://developers.googleblog.com/2017/12/creating-custom-estimators-in-tensorflow.html

* Latest (Official) MNIST using customer estimator:
https://github.com/tensorflow/models/blob/master/official/mnist/mnist.py

Talk 2: Deep Dive into Google's TPUs including their new Pipeline Profiling Tools within TensorBoard (by Romit Singhai, Applied AI Engineer @ GE)

https://www.linkedin.com/in/romitsinghai/

Romit will share his experience working with Google's new TPUs on various datasets, training algorithms, and device-placement strategies.

He will highlight the differences between the standard CPU/GPU Estimator API - and the new TPU Estimator API. These are important differences, so pay close attention!

In addition, Romit and I discovered some new TensorBoard profiling features that analyze your ENTIRE TensorFlow pipeline including data ingestion and ETL to CPU, GPU, and TPU utilization and graph/operator optimization.

Note: These profiling tools are exactly what we've always from Spark-based ETL pipelines, but we've never seen them on the market - not at this level of system detail and optimization.

Lastly, Romit will perform a live demo of TPU training, profiling, and optimizing - complete with source code and runtime configuration.

Talk 3: The Math Behind Neural Networks by Francesco Mosconi, PhD (https://www.linkedin.com/in/framosconis/)

Francesco is Founder and Data Scientist @ CATALIT Data Science - a Deep Learning and Advanced Analytics Consultancy/Training Company based in San Francisco.

Francesco is also Founder of the popular Data Weekends: https://www.dataweekends.com/

Related Links
* https://pipeline.ai
* https://youtube.pipeline.ai
* https://slideshare.pipeline.ai
* https://quickstart.pipeline.ai
* http://community.pipeline.ai
* https://www.neuraldesigner.com/blog/5_algorithms_to_train_a_neural_network
* http://algorithms-tour.stitchfix.com/

PLEASE RSVP HERE TO MAKE SURE YOU GET A SPOT:

https://www.eventbrite.com/e/deep-dive-into-google-tpu-tfrecord-dataset-api-kafka-math-behind-neural-nets-tickets-44304797843