runtime

runtime copied to clipboard

Work stealing method in **NonBlockingWorkQueue** is not a good idea to reduce cache missing in LLC automatically. Depth-first execution computation graphs are generally more cache-friendly. If I want to apply...

MoFHeka

TFRT support for Mobiles(Android) or Bare Metal

1

Hello, Does TFRT supports Mobiles(Android) or Bare Metal Embedded Systems? Do you have example for running TFRT on Mobiles(Android) or Bare Metal Embedded Systems?

uccoder-hkm

There is a failure while trying to build TensorFlow-CPU on the Windows platform on the CLANG compiler. The detailed error log is in the link below https://tensorflow-ci.intel.com/view/Windows/job/tf-clang/47/artifact/test_run.log

mraunak

Mobile PoC level support ETA

4

When is the ETA for supporting mobile devices ? I wonder the expected performance and memory footprint on android devices. Thank you.

chunseoklee

cetinken

is it possible to support dynamic shape from the design of TFRT? if so, could you please share the idea ? many thanks

FullZing

Otherwise getting `ERROR: error loading package '': Label '//cuda:dependencies.bzl' is invalid because 'cuda' is not a package; perhaps you meant to put the colon here: '//:cuda/dependencies.bzl'?` Thanks for your contribution!...

REASY

Fold rules_cuda into bazelbuild/rules_cuda

4

Ideally `rules_cuda` should be split away from this project. Even the README of `rules_cuda` has the repo located at its own repository. You probably have this tracked internally, but for...

dmadisetti

When I use ResNet50 with XLA enabled in tensorflow with a simple file to benchmark XLA performance under different batch sizes, but the model can run successfully in batch size...

AmosChenYQ

the dependency version does not seem to be up to date

1

hi, it seems that the dependency version(e.g. bazel, clang) in your readme is not up to date, can you show the newest version to build TFRT？

lethean1