experiments/STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Remove NCCL from RBE dockers. NCCL is built from source now.

Browse Source 
PiperOrigin-RevId: 227661008
This commit is contained in:
A. Unique TensorFlower 2019-01-03 03:11:44 -08:00 committed by TensorFlower Gardener
parent 37326d1497
commit eb54349cb4
3 changed files with 0 additions and 28 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
tensorflow/tools/ci_build/Dockerfile.rbe.cuda10.0-cudnn7-ubuntu14.04
View File

@ -19,7 +19,6 @@ RUN apt-get update && apt-get install -y --no-install-recommends ca-certificates
ENV CUDA_VERSION 10.0.130 ENV CUDA_VERSION 10.0.130
ENV CUDA_PKG_VERSION 10-0=$CUDA_VERSION-1 ENV CUDA_PKG_VERSION 10-0=$CUDA_VERSION-1
ENV CUDNN_VERSION 7.3.1.20 ENV CUDNN_VERSION 7.3.1.20
ENV NCCL_VERSION 2.3.5
ENV TENSORRT_VERSION 5.0.2 ENV TENSORRT_VERSION 5.0.2
ENV NVIDIA_DRIVER_CAPABILITIES compute,utility ENV NVIDIA_DRIVER_CAPABILITIES compute,utility
ENV NVIDIA_REQUIRE_CUDA "cuda>=10.0,driver>=410" ENV NVIDIA_REQUIRE_CUDA "cuda>=10.0,driver>=410"
@ -48,26 +47,18 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
libcudnn7=$CUDNN_VERSION-1+cuda10.0 \ libcudnn7=$CUDNN_VERSION-1+cuda10.0 \
libcudnn7=$CUDNN_VERSION-1+cuda10.0 \ libcudnn7=$CUDNN_VERSION-1+cuda10.0 \
libcudnn7-dev=$CUDNN_VERSION-1+cuda10.0 \ libcudnn7-dev=$CUDNN_VERSION-1+cuda10.0 \
libnccl2=$NCCL_VERSION-2+cuda10.0 \
libnccl-dev=$NCCL_VERSION-2+cuda10.0 \
nvinfer-runtime-trt-repo-ubuntu1604-$TENSORRT_VERSION-ga-cuda10.0 && \ nvinfer-runtime-trt-repo-ubuntu1604-$TENSORRT_VERSION-ga-cuda10.0 && \
apt-get update && apt-get install -y --no-install-recommends \ apt-get update && apt-get install -y --no-install-recommends \
libnvinfer5=$TENSORRT_VERSION-1+cuda10.0 \ libnvinfer5=$TENSORRT_VERSION-1+cuda10.0 \
libnvinfer-dev=$TENSORRT_VERSION-1+cuda10.0 && \ libnvinfer-dev=$TENSORRT_VERSION-1+cuda10.0 && \
ln -s cuda-10.0 /usr/local/cuda && \ ln -s cuda-10.0 /usr/local/cuda && \
apt-mark hold libcudnn7 && \ apt-mark hold libcudnn7 && \
apt-mark hold libnccl2 && \
rm -rf /var/lib/apt/lists/* rm -rf /var/lib/apt/lists/*
# TODO(b/110903506): Provide a link to the SONAME of libcuda.so. # TODO(b/110903506): Provide a link to the SONAME of libcuda.so.
# https://github.com/NVIDIA/nvidia-docker/issues/775 # https://github.com/NVIDIA/nvidia-docker/issues/775
RUN ln -s libcuda.so /usr/local/cuda/lib64/stubs/libcuda.so.1 RUN ln -s libcuda.so /usr/local/cuda/lib64/stubs/libcuda.so.1
# TODO(klimek): Once the TODO in tensorflow's configure.py to correctly find
# libnccl is resolved, delete this block.
RUN ln -s /usr/lib/x86_64-linux-gnu/libnccl.so /usr/lib/libnccl.so \
&& ln -s /usr/lib/x86_64-linux-gnu/libnccl.so /usr/lib/libnccl.so.2
# Install a newer version of libstdc++, as new clang versions do not work # Install a newer version of libstdc++, as new clang versions do not work
# with the stock ubuntu 14.04 libstdc++. # with the stock ubuntu 14.04 libstdc++.
RUN apt-get update && \ RUN apt-get update && \

9
tensorflow/tools/ci_build/Dockerfile.rbe.cuda9.0-cudnn7-ubuntu14.04
View File

@ -25,7 +25,6 @@ ENV PATH /usr/local/nvidia/bin:/usr/local/cuda/bin:${PATH}
ENV NVIDIA_VISIBLE_DEVICES all ENV NVIDIA_VISIBLE_DEVICES all
ENV NVIDIA_DRIVER_CAPABILITIES compute,utility ENV NVIDIA_DRIVER_CAPABILITIES compute,utility
ENV NVIDIA_REQUIRE_CUDA "cuda>=9.0" ENV NVIDIA_REQUIRE_CUDA "cuda>=9.0"
ENV NCCL_VERSION 2.2.13
ENV TENSORRT_VERSION 5.0.2 ENV TENSORRT_VERSION 5.0.2
ENV CUDNN_VERSION 7.1.4.18 ENV CUDNN_VERSION 7.1.4.18
@ -45,14 +44,12 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
cuda-cudart-$CUDA_PKG_VERSION \ cuda-cudart-$CUDA_PKG_VERSION \
cuda-libraries-$CUDA_PKG_VERSION \ cuda-libraries-$CUDA_PKG_VERSION \
cuda-cublas-9-0=9.0.176.4-1 \ cuda-cublas-9-0=9.0.176.4-1 \
libnccl2=$NCCL_VERSION-1+cuda9.0 \
cuda-libraries-dev-$CUDA_PKG_VERSION \ cuda-libraries-dev-$CUDA_PKG_VERSION \
cuda-nvml-dev-$CUDA_PKG_VERSION \ cuda-nvml-dev-$CUDA_PKG_VERSION \
cuda-minimal-build-$CUDA_PKG_VERSION \ cuda-minimal-build-$CUDA_PKG_VERSION \
cuda-command-line-tools-$CUDA_PKG_VERSION \ cuda-command-line-tools-$CUDA_PKG_VERSION \
cuda-core-9-0=9.0.176.3-1 \ cuda-core-9-0=9.0.176.3-1 \
cuda-cublas-dev-9-0=9.0.176.4-1 \ cuda-cublas-dev-9-0=9.0.176.4-1 \
libnccl-dev=$NCCL_VERSION-1+cuda9.0 \
libcudnn7-dev=$CUDNN_VERSION-1+cuda9.0 \ libcudnn7-dev=$CUDNN_VERSION-1+cuda9.0 \
libcudnn7=$CUDNN_VERSION-1+cuda9.0 \ libcudnn7=$CUDNN_VERSION-1+cuda9.0 \
nvinfer-runtime-trt-repo-ubuntu1604-$TENSORRT_VERSION-ga-cuda9.0 && \ nvinfer-runtime-trt-repo-ubuntu1604-$TENSORRT_VERSION-ga-cuda9.0 && \
@ -60,7 +57,6 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
libnvinfer5=$TENSORRT_VERSION-1+cuda9.0 \ libnvinfer5=$TENSORRT_VERSION-1+cuda9.0 \
libnvinfer-dev=$TENSORRT_VERSION-1+cuda9.0 && \ libnvinfer-dev=$TENSORRT_VERSION-1+cuda9.0 && \
ln -s cuda-9.0 /usr/local/cuda && \ ln -s cuda-9.0 /usr/local/cuda && \
apt-mark hold libnccl2 && \
apt-mark hold libcudnn7 libcudnn7-dev && \ apt-mark hold libcudnn7 libcudnn7-dev && \
rm -rf /var/lib/apt/lists/* rm -rf /var/lib/apt/lists/*
@ -71,11 +67,6 @@ RUN echo "/usr/local/nvidia/lib" >> /etc/ld.so.conf.d/nvidia.conf && \
# https://github.com/NVIDIA/nvidia-docker/issues/775 # https://github.com/NVIDIA/nvidia-docker/issues/775
RUN ln -s libcuda.so /usr/local/cuda/lib64/stubs/libcuda.so.1 RUN ln -s libcuda.so /usr/local/cuda/lib64/stubs/libcuda.so.1
# TODO(klimek): Once the TODO in tensorflow's configure.py to correctly find
# libnccl is resolved, delete this block.
RUN ln -s /usr/lib/x86_64-linux-gnu/libnccl.so /usr/lib/libnccl.so \
&& ln -s /usr/lib/x86_64-linux-gnu/libnccl.so /usr/lib/libnccl.so.2
# Install a newer version of libstdc++, as new clang versions do not work # Install a newer version of libstdc++, as new clang versions do not work
# with the stock ubuntu 14.04 libstdc++. # with the stock ubuntu 14.04 libstdc++.
RUN apt-get update && \ RUN apt-get update && \

10
tensorflow/tools/docker/Dockerfile.devel-gpu
View File

@ -15,8 +15,6 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
git \ git \
libcudnn7=7.2.1.38-1+cuda9.0 \ libcudnn7=7.2.1.38-1+cuda9.0 \
libcudnn7-dev=7.2.1.38-1+cuda9.0 \ libcudnn7-dev=7.2.1.38-1+cuda9.0 \
libnccl2=2.2.13-1+cuda9.0 \
libnccl-dev=2.2.13-1+cuda9.0 \
libcurl3-dev \ libcurl3-dev \
libfreetype6-dev \ libfreetype6-dev \
libhdf5-serial-dev \ libhdf5-serial-dev \
@ -41,11 +39,6 @@ RUN apt-get update && \
apt-get install libnvinfer4=4.1.2-1+cuda9.0 && \ apt-get install libnvinfer4=4.1.2-1+cuda9.0 && \
apt-get install libnvinfer-dev=4.1.2-1+cuda9.0 apt-get install libnvinfer-dev=4.1.2-1+cuda9.0
# Link NCCL libray and header where the build script expects them.
RUN mkdir /usr/local/cuda-9.0/lib && \
ln -s /usr/lib/x86_64-linux-gnu/libnccl.so.2 /usr/local/cuda/lib/libnccl.so.2 && \
ln -s /usr/include/nccl.h /usr/local/cuda/include/nccl.h
RUN curl -fSsL -O https://bootstrap.pypa.io/get-pip.py && \ RUN curl -fSsL -O https://bootstrap.pypa.io/get-pip.py && \
python get-pip.py && \ python get-pip.py && \
rm get-pip.py rm get-pip.py
@ -111,9 +104,6 @@ ENV TF_CUDA_COMPUTE_CAPABILITIES=3.5,5.2,6.0,6.1,7.0
ENV TF_CUDA_VERSION=9.0 ENV TF_CUDA_VERSION=9.0
ENV TF_CUDNN_VERSION=7 ENV TF_CUDNN_VERSION=7
# NCCL 2.x
ENV TF_NCCL_VERSION=2
RUN ln -s /usr/local/cuda/lib64/stubs/libcuda.so /usr/local/cuda/lib64/stubs/libcuda.so.1 && \ RUN ln -s /usr/local/cuda/lib64/stubs/libcuda.so /usr/local/cuda/lib64/stubs/libcuda.so.1 && \
LD_LIBRARY_PATH=/usr/local/cuda/lib64/stubs:${LD_LIBRARY_PATH} \ LD_LIBRARY_PATH=/usr/local/cuda/lib64/stubs:${LD_LIBRARY_PATH} \
tensorflow/tools/ci_build/builds/configured GPU \ tensorflow/tools/ci_build/builds/configured GPU \