TensorFlow 1.13.1, GPU-skylake (CUDA 10.0, cuDNN 7.5, TensorRT 5; MKL) Py37, Buntu 18.04 (x64)

TF	Py	CUDA	cuDNN	TensorRT	AditionalOpts	Opt Flags
1.13.1	3.7	10.0	7.5	5.0	✔️ XLA JIT ✔️ MKL	* SSE4.1, 4.2 * AVX, AVX2 * FMA

• Compiled on:
  • 4.15.0-45
  • Xubuntu 18.04
  • GCC-7
  • -march=skylake
• Extras:
  • XLA JIT
  • mkl
  • nccl 2.4.2

Requires:

• All versions and libraries featured
  • CUDA 10
  • cuDNN 7.5
  • TRT 5
  • Buntu 18.04 OR x64 linux with glibc 2.27 (UNTESTED)
• Nvidia GPU (this wheel optimized for CUDA compute capability 6.1)
• Broadwell or later Intel CPU, (this wheel optimized for -march=skylake

Installation

From downloaded wheel

pip install --no-cache-dir tensorflow-1.13.1-cp37-cp37m-linux_x86_64.whl

From wheel link

pip install --no-cache-dir https://github.com/evdcush/TensorFlow-wheels/releases/download/tf-1.13.1-py37-gpu-10.0/tensorflow-1.13.1-cp37-cp37m-linux_x86_64.whl

TensorFlow 1.13.1, CPU-westmere (SSE4.1, SSE4.2; MKL), Py37, Buntu 18.04 (x64)

TF	Buntu	Py	march	SSE4.1	SSE4.2	AVX	AVX2	FMA	MKL
1.13.1	18.04	3.7.2	westmere	✔️	✔️	❌	❌	❌	✔️

• Compiled on:
  • 4.15.0-45
  • Xubuntu 18.04 (amd64)
  • GCC-7
  • -march=westmere
• Extras:
  • intel mkl

Requires:

• Westmere or newer Intel CPU
• Buntu >= 18.04, OR x64 linux with glibc 2.27

Installation

From downloaded wheel

pip install --no-cache-dir tensorflow-1.13.1-cp37-cp37m-linux_x86_64.whl

From wheel link

pip install --no-cache-dir https://github.com/evdcush/TensorFlow-wheels/releases/download/tf-1.13.1-py37-cpu-westmere/tensorflow-1.13.1-cp37-cp37m-linux_x86_64.whl

TensorFlow 1.13.1, CPU-ivybridge (SSE4.1, SSE4.2, AVX; MKL), Py37, Buntu 18.04 (x64)

TF	Buntu	Py	march	SSE4.1	SSE4.2	AVX	AVX2	FMA	MKL
1.13.1	18.04	3.7.2	ivybridge	✔️	✔️	✔️	❌	❌	✔️

• Compiled on:
  • 4.15.0-45
  • Xubuntu 18.04 (amd64)
  • GCC-7
  • -march=ivybridge
• Extras:
  • intel mkl

Requires:

• Ivy Bridge or newer Intel CPU
• Buntu >= 18.04, OR x64 linux with glibc 2.27

Installation

From downloaded wheel

pip install --no-cache-dir tensorflow-1.13.1-cp37-cp37m-linux_x86_64.whl

From wheel link

pip install --no-cache-dir https://github.com/evdcush/TensorFlow-wheels/releases/download/tf-1.13.1-py37-cpu-ivybridge/tensorflow-1.13.1-cp37-cp37m-linux_x86_64.whl

TensorFlow 1.12.0, Py37, CPU-westmere (SSE4.1, SSE4.2), Buntu 18.04

TF	Buntu	Py	march	SSE4.1	SSE4.2	AVX	AVX2	FMA	MKL
1.12	18.04	3.7.2	westmere	✔️	✔️	❌	❌	❌	❌

• Compiled on: 4.15.0-43, Xubuntu 18.04 (amd64), GCC-7 with -march=westmere
• Extras:

Requires:

• Westmere or newer Intel CPU
• Buntu >= 18.04, OR x64 linux with glibc 2.27

Installation

From downloaded wheel

pip install --no-cache-dir tensorflow-1.12.0-cp37-cp37m-linux_x86_64.whl

From wheel link

pip install --no-cache-dir https://github.com/evdcush/TensorFlow-wheels/releases/download/tf-1.12.0-py37-cpu-westmere/tensorflow-1.12.0-cp37-cp37m-linux_x86_64.whl

TensorFlow 1.12.0, Py37, CPU-ivybridge (SSE4.1, SSE4.2, AVX; MKL), Buntu 16.04

TF	Buntu	Py	march	SSE4.1	SSE4.2	AVX	AVX2	FMA	MKL
1.12	16.04	3.7.1	ivybridge	✔️	✔️	✔️	❌	❌	✔️

• Compiled on: 4.15.0-39, Xubuntu 16.04 (amd64), GCC-7 with -march=ivybridge
• Extras: MKL

Requires:

• Ivy Bridge or newer Intel CPU
• Buntu >= 16.04, OR x64 linux with glibc 2.23 (UNTESTED)

Installation

From downloaded wheel

pip install --no-cache-dir tensorflow-1.12.0-cp37-cp37m-linux_x86_64.whl

From wheel link

pip install --no-cache-dir https://github.com/evdcush/TensorFlow-wheels/releases/download/tf-1.12.0-py37-cpu-ivybridge/tensorflow-1.12.0-cp37-cp37m-linux_x86_64.whl

TensorFlow 1.12, GPU-skylake (CUDA 10.0, cuDNN 7.3, TensorRT 5; MKL) Py37, Buntu 18.04 (x64)

TF	Py	CUDA	cuDNN	TensorRT	AditionalOpts	Opt Flags
1.12.0	3.7	10.0	7.3	5.0	✔️ XLA JIT ✔️ MKL	* SSE4.1, 4.2 * AVX, AVX2 * FMA

• Compiled on:
  • 4.15.0-38
  • Xubuntu 18.04 (x64)
  • GCC-7
  • -march=skylake
• Extras:
  • XLA JIT
  • mkl
  • nccl 2.3

Requires:

• All versions and libraries featured
  • CUDA 10
  • cuDNN 7.3
  • TRT 5
  • Buntu 18.04 OR x64 linux with glibc 2.27 (UNTESTED)
• Nvidia GPU (this wheel optimized for CUDA compute capability 6.1)
• Broadwell or later Intel CPU, (this wheel optimized for -march=skylake

Installation

From downloaded wheel

pip install --no-cache-dir tensorflow-1.12.0-cp37-cp37m-linux_x86_64.whl

From wheel link

pip install --no-cache-dir https://github.com/evdcush/TensorFlow-wheels/releases/download/tf-1.12.0-py37-gpu-10.0/tensorflow-1.12.0-cp37-cp37m-linux_x86_64.whl

TensorFlow 1.12.0 GPU (CUDA 10.0, cuDNN 7.3, TensorRT 5; MKL), Py36, Buntu 18.04

UPDATED: binary

The original release binary had issues ngraph (core dumps 💥), so I rebuilt the wheel without ngraph.

TF	Py	CUDA	cuDNN	TensorRT	AditionalOpts	Opt Flags
1.12	3.6.7	10.0	7.3	5.0	✔️ XLA JIT ✔️ MKL	* SSE4.1, 4.2 * AVX, AVX2 * FMA

• Compiled on: 4.15.0-38, Xubuntu 18.04, GCC-7 with -march=skylake
• Extras: XLA JIT (now supported by default in 1.12), mkl, nccl 2.3

Requires:

• All versions and libraries featured
  • CUDA 10, cuDNN 7.3, TRT 5, Buntu 18.04 OR x64 linux with glibc 2.27 (UNTESTED)
• Nvidia GPU (this wheel optimized for CUDA compute capability 6.1)
• Broadwell or later Intel CPU, (this wheel optimized for -march=skylake

Installation

From downloaded wheel

pip install --no-cache-dir tensorflow-1.12.0-cp36-cp36m-linux_x86_64.whl

From wheel link

pip install --no-cache-dir https://github.com/evdcush/TensorFlow-wheels/releases/download/tf-1.12.0-gpu-10.0/tensorflow-1.12.0-cp36-cp36m-linux_x86_64.whl

TensorFlow 1.11.0 GPU (CUDA 10.0, cuDNN 7.3, TensorRT 5, MKL)

TF	OS	Py	CUDA	cuDNN	TensorRT	MKL	Opt Flags
1.11.0	Linux	3.6	10.0	7.3	5.0	✔️	* SSE4.1, 4.2 * AVX, AVX2 * FMA

• Compiled on: 4.15.0-36, Xubuntu 18.04, GCC-7 with -march=skylake
• Extras: XLA JIT and nothing else (nGraph, NCCL 2.3 unsupported in 1.11.0 apparently)

Requires:

• All versions and libraries featured in columns
• Nvidia GPU (this wheel optimized for CUDA compute capability 6.1)
• Broadwell or later Intel CPU, (this wheel optimized for -march=skylake)

TensorFlow 1.10.0 GPU (CUDA 9.2, cuDNN 7.2, TensorRT 4, MKL)

Version	OS	Py	CUDA	cuDNN	TensorRT	MKL	Opt Flags
1.10.0	Linux	3.6	9.2	7.2	✔️	✔️	SSE4.1 SSE4.2 AVX AVX2 FMA

Requires:

• All versions and libraries featured in columns
• Nvidia GPU (this wheel optimized for CUDA compute capability 6.1)
• Broadwell or later Intel CPU, (this wheel optimized for -march=skylake)

TensorFlow 1.10.0 GPU (CUDA 9.1, cuDNN 7.1, MKL)

Version	OS	Py	CUDA	cuDNN	TensorRT	MKL	Opt Flags
1.10.0	Linux	3.6	9.1	7.1	❌	✔️	SSE4.1 SSE4.2 AVX AVX2 FMA

Requires:

• All versions and libraries featured in columns
• Nvidia GPU (this wheel optimized for CUDA compute capability 6.1)
• Broadwell or later Intel CPU, (this wheel optimized for -march=broadwell)