Home

Awesome

This repository contains custom builds of tensorflow. To install one of these on your system, download the correct file according to your version of python and gcc and run the following command.

pip install --ignore-installed --upgrade /path/to/binary.whl --user

You can also install directly from github using

pip install --ignore-installed --upgrade "Download URL" --user

All the binaries were compiled by passing --config=opt to bazel and setting -O2 -Wno-sign-compare in addition to the feature flags in configure script.

Latest Builds

TFHWOSGCCPythonNumPySupports
2.7.0CPUUbuntu 20.049.3.03.8.101.21.4FMA, AVX, AVX2, SSE4.1, SSE4.2Download
2.2.1CPUmacOS Catalinaclang-1200.0.26.23.8.6FMA, AVX, AVX2, SSE4.1, SSE4.2Download
2.2.0CPUUbuntu 18.047.5.03.7.7FMA, AVX, AVX2, SSE4.1, SSE4.2Download
2.1.0CPUUbuntu 16.045.4.03.5.2FMA, AVX, AVX2, SSE4.1, SSE4.2Download

What flags are supported by my CPU?

To check the instruction sets supported by your CPU, check the output of the following commands:

# macOS
sysctl -a | grep "machdep.cpu.*features:"

# Linux
cat /proc/cpuinfo | grep flags
<details> <summary>Expand for older builds</summary>
TFHWOSGCCPythonSupports
2.4.1CPUUbuntu 20.049.3.03.8.5FMA, AVX, AVX2, SSE4.1, SSE4.2Download
1.14.1CPUmacOS Mojaveclang-1001.0.46.43.7.4FMA, AVX, AVX2, SSE4.1, SSE4.2Download
1.14.1CPUmacOS Mojaveclang-1001.0.46.43.7.4FMA, AVX2, AVX512FDownload
1.14.1CPUUbuntu 18.047.43.7.3FMA, AVX2, AVX512FDownload
1.13.1CPUMacOS Mojaveclang-1001.0.46.43.7.2FMA, AVX, AVX2, SSE4.1, SSE4.2Download
1.13.1CPUUbuntu 16.045.43.5.2FMA, AVX, AVX2, SSE4.1, SSE4.2Download
1.12.0CPUUbuntu 16.045.43.5.4FMA, AVX2, AVX512FDownload
1.12.0CPUUbuntu 18.047.33.6.5FMA, AVX2, AVX512FDownload
1.12.0CPUUbuntu 18.047.33.6.5FMA, AVX, AVX2, SSE4.1, SSE4.2Download
1.12.0CPUUbuntu 18.047.32.7.15rc1FMA, AVX, AVX2, SSE4.1, SSE4.2Download
1.12.0CPUUbuntu 16.045.43.5.2FMA, AVX, AVX2, SSE4.1, SSE4.2Download
1.12.0CPUUbuntu 16.045.42.7.12FMA, AVX, AVX2, SSE4.1, SSE4.2Download
1.12.0CPUMacOS Mojaveclang-1000.10.44.43.7.0FMA, AVX, AVX2, SSE4.1, SSE4.2Download
1.12.0CPUMacOS Mojaveclang-1000.10.44.42.7.15FMA, AVX, AVX2, SSE4.1, SSE4.2Download
1.11.0CPUUbuntu 16.045.43.5.2FMA, AVX, AVX2, SSE4.1, SSE4.2Download
1.11.0CPUUbuntu 16.045.42.7.12FMA, AVX, AVX2, SSE4.1, SSE4.2Download
1.11.0CPUMacOS Mojaveclang-1000.10.44.23.7.0FMA, AVX, AVX2, SSE4.1, SSE4.2Download
1.11.0CPUMacOS Mojaveclang-1000.10.44.22.7.15FMA, AVX, AVX2, SSE4.1, SSE4.2Download
1.10.0CPUUbuntu 16.045.43.6.6FMA, AVX, AVX2, SSE4.1, SSE4.2, AVX512FDownload
1.10.0CPUUbuntu 16.045.42.7.12FMA, AVX, AVX2, SSE4.1, SSE4.2Download
1.10.0CPUUbuntu 16.045.43.5.2FMA, AVX, AVX2, SSE4.1, SSE4.2Download
1.10.0CPUUbuntu 18.047.32.7.15rc1FMA, AVX, AVX2, SSE4.1, SSE4.2Download
1.10.0CPUUbuntu 18.047.33.6.5FMA, AVX, AVX2, SSE4.1, SSE4.2Download
1.9.0CPUUbuntu 18.047.33.6.5FMA, AVX, AVX2, SSE4.1, SSE4.2Download
1.9.0CPUUbuntu 16.045.42.7.12FMA, AVX, AVX2, SSE4.1, SSE4.2Download
1.9.0CPUUbuntu 16.045.43.5.2FMA, AVX, AVX2, SSE4.1, SSE4.2Download
1.9.0CPUUbuntu 16.045.43.6.6FMA, AVX, AVX2, SSE4.1, SSE4.2Download
1.9.0CPUmacOS High Sierraclang-902.0.39.12.7.15SSE4.1, SSE4.2, AVX, AVX2, FMADownload
1.9.0CPUmacOS High Sierraclang-902.0.39.13.6.5SSE4.1, SSE4.2, AVX, AVX2, FMADownload
1.8.0CPUmacOS High Sierraclang-902.0.39.12.7.15SSE4.1, SSE4.2, AVX, AVX2, FMADownload
1.8.0CPUmacOS High Sierraclang-902.0.39.13.6.5SSE4.1, SSE4.2, AVX, AVX2, FMADownload
1.8.0CPUmacOS High Sierraclang-902.0.39.13.6.5SSE4.1, SSE4.2, AVX, AVX2, FMA, SSSE3, POPCNT, CX16Download
1.6.0CPUUbuntu 16.045.42.7.12SSE4.1, SSE4.2, AVX, AVX2, FMADownload
1.5.0CPUmacOS High Sierraclang-900.0.39.23.6.4AVX, SSE4.1, SSE4.2Download
1.5.0CPUmacOS High Sierraclang-900.0.39.23.6.4SSE4.2, AVX, AVX2, FMADownload
1.4.1CPUmacOS Sierraclang-900.0.39.23.6.1AVX, SSE4.1, SSE4.2Download
1.4.1CPUmacOS Sierraclang-900.0.39.22.7.13AVX, SSE4.1, SSE4.2Download
1.4.1CPUmacOS Sierraclang-900.0.39.22.7.14SSE4.1, SSE4.2, AVX, AVX2, FMADownload
1.4.0CPUmacOS Sierraclang-802.0.423.6.3SSE4.1, SSE4.2, AVX, AVX2, FMADownload
</details>

Visit tensorflow.org for instructions on building tensorflow for your machine.