TensorFlow 1.7 (GPU, CUDA 9.1, cuDNN 7.1, no MKL)

GPU wheels for TensorFlow 1.7, built by TinyMind, the cloud machine learning platform.

To use the wheels on your own machine, Intel Broadwell or later CPU, and Nvidia computing capability 3.7 or later GPU with CUDA 9.1 and cuDNN 7.1 are required.

This version is optimized for compute capabilities 3.7 (K80, AWS P2/GCP), 6.0 (P100, GCP) and 7.0 (V100, AWS P3).

TensorFlow 1.7 (CPU, MKL)

CPU wheels for TensorFlow 1.7 with MKL. You may want to take a look at the performance guide for MKL.

NOTE: These wheels contain MKL support. If you don't have it, install MKL by following the instructions here.

You can find the non-MKL CPU version for TensorFlow 1.7 here.

TensorFlow 1.7 (CPU)

CPU wheels for TensorFlow 1.7, built by TinyMind, the cloud machine learning platform.

To use the wheels on your own machine, Intel Broadwell or later CPU is required.

TensorFlow 1.6 (GPU, CUDA 9.1, no MKL)

GPU wheels for TensorFlow 1.6, built by TinyMind, the cloud machine learning platform.

To use the wheels on your own machine, Intel Broadwell or later CPU, and Nvidia computing capability 3.7 or later GPU with CUDA 9.1 and CuDNN 7.1.2 are required.

This version is optimized for compute capabilities 3.7 (K80, AWS P2/GCP).

TensorFlow 1.6 (GPU, CUDA 9.1)

GPU wheels for TensorFlow 1.6, built by TinyMind, the cloud machine learning platform.

To use the wheels on your own machine, Intel Broadwell or later CPU, and Nvidia computing capability 3.7 or later GPU with CUDA 9.1 and CuDNN 7 are required.

This version is optimized for compute capabilities 3.7 (K80, AWS P2/GCP), 6.0 (P100, GCP) and 7.0 (V100, AWS P3).

NOTE: These wheels contain MKL support. If you don't have it, install MKL by following the instructions here. You may want to take a look at the performance guide for MKL.

TensorFlow 1.6 (CPU, MKL)

CPU wheels for TensorFlow 1.6 with MKL. You may want to take a look at the performance guide for MKL.

NOTE: These wheels contain MKL support. If you don't have it, install MKL by following the instructions here.

You can find the non-MKL CPU version for TensorFlow 1.6 here.

TensorFlow 1.6 (CPU)

CPU wheels for TensorFlow 1.6, built by TinyMind, the cloud machine learning platform.

To use the wheels on your own machine, Intel Broadwell or later CPU is required.

TensorFlow Serving 1.5

TensorFlow Serving 1.5 with Python API. The model-server.gz file contains bazel build output for the server and can be put at the root level of the cloned tensorflow-serving folder as bazel-bin.

TensorFlow 1.5 (GPU, CUDA 9, no MKL)

GPU wheels for TensorFlow 1.5, built by TinyMind, the cloud machine learning platform.

To use the wheels on your own machine, Intel Broadwell or later CPU, and Nvidia computing capability 3.7 or later GPU with CUDA 9 and CuDNN 7 are required.

This version is optimized for compute capabilities 3.7 (K80, AWS P2/GCP), 6.0 (P100, GCP) and 7.0 (V100, AWS P3).

TensorFlow 1.5 (GPU, CUDA 9.1, Generic)

CUDA 9.1 wheels for TensorFlow 1.5, without optimization. We provide them here as TensorFlow hasn't officially provided CUDA 9.1 wheels, and our optimized builds may not work for AMD or old Intel CPUs.