This course provides an inside look into the TensorFlow team's internal training sessions, focusing on technical deep dives into TensorFlow by the developers themselves. The course covers topics such as eager execution runtime, including defining and generating code, Python context, execution modes, kernel operations, code generation, function calls, multi-device operations, TensorFlow Executor, and memory management. The teaching method involves a presentation by Software Engineer Alex Passos. This course is intended for individuals interested in gaining a deeper understanding of TensorFlow's eager execution runtime.
Overview
Syllabus
Introduction
Defining reloj
Generating reloj
Python context
Python vs C
Eager code
tensor handle
execution
Execution in Python
Eager execute
Sync mode
Calling a kernel
Class kernel
Kernels
Code generation
Summary
Async mode
Function calls
Function library runtime
Multidevice call
Rendezvous
Tensorflow Executor
Highlights
Host vs device memory
TF in 32 problem
Solution
Taught by
TensorFlow
