PyTorch 1.2.0 (GPU, CUDA 9.1.85, CuDNN 7.1.3.16, NNPACK)
·
1 commit
to master
since this release
GPU wheels for PyTorch 1.2.0, built by Rosto.io.
To use the wheels on your own machine Nvidia computing capability 3.5 or later GPU with CUDA 9.1.85 and cuDNN 7.1.3 are required.
Compilation flags TORCH_NVCC_FLAGS="-Xfatbin -compress-all".
This version is optimized for compute capabilities 3.5, 5.2, 6.0, 6.1 and 7.0. It includes, but not limited to, GeForce GTX Titan Z, GTX Titan, GTX 780, GT 730, GT 720, GTX Titan X, 980 Ti, 970, 960, 950, 750 SE, 980M, 970M, 965M, P100, GCP, GTX1050, 1060, 1070, 1080, 1080Ti, V100, AWS P3.
NNPack is supported.