Upgrade to CUDA 10.1

Author: rosbo

Dockerfile

@@ -1,7 +1,7 @@
 ARG BASE_TAG=2019.03
 ARG TENSORFLOW_VERSION=2.1.0
 
-FROM gcr.io/kaggle-images/python-tensorflow-whl:${TENSORFLOW_VERSION}-py36 as tensorflow_whl
+FROM gcr.io/kaggle-images/python-tensorflow-whl:${TENSORFLOW_VERSION}-py36-2 as tensorflow_whl
 FROM continuumio/anaconda3:${BASE_TAG}
 
 ADD clean-layer.sh  /tmp/clean-layer.sh

gpu.Dockerfile

@@ -1,7 +1,7 @@
 ARG BASE_TAG=staging
 
-FROM nvidia/cuda:10.0-cudnn7-devel-ubuntu16.04 AS nvidia
-FROM gcr.io/kaggle-images/python-tensorflow-whl:2.1.0-py36 as tensorflow_whl
+FROM nvidia/cuda:10.1-cudnn7-devel-ubuntu16.04 AS nvidia
+FROM gcr.io/kaggle-images/python-tensorflow-whl:2.1.0-py36-2 as tensorflow_whl
 FROM gcr.io/kaggle-images/python:${BASE_TAG}
 
 ADD clean-layer.sh  /tmp/clean-layer.sh
@@ -13,8 +13,11 @@ COPY --from=nvidia /etc/apt/trusted.gpg /etc/apt/trusted.gpg.d/cuda.gpg
 
 # Ensure the cuda libraries are compatible with the custom Tensorflow wheels.
 # TODO(b/120050292): Use templating to keep in sync or COPY installed binaries from it.
-ENV CUDA_VERSION=10.0.130
-ENV CUDA_PKG_VERSION=10-0=$CUDA_VERSION-1
+ENV CUDA_MAJOR_VERSION=10
+ENV CUDA_MINOR_VERSION=1
+ENV CUDA_PATCH_VERSION=243
+ENV CUDA_VERSION=$CUDA_MAJOR_VERSION.$CUDA_MINOR_VERSION.$CUDA_PATCH_VERSION
+ENV CUDA_PKG_VERSION=$CUDA_MAJOR_VERSION-$CUDA_MINOR_VERSION=$CUDA_VERSION-1
 LABEL com.nvidia.volumes.needed="nvidia_driver"
 LABEL com.nvidia.cuda.version="${CUDA_VERSION}"
 ENV PATH=/usr/local/nvidia/bin:/usr/local/cuda/bin:${PATH}
@@ -26,7 +29,7 @@ ENV PATH=/usr/local/nvidia/bin:/usr/local/cuda/bin:${PATH}
 ENV LD_LIBRARY_PATH="/usr/local/nvidia/lib64:/usr/local/cuda/lib64:/usr/local/cuda/lib64/stubs"
 ENV NVIDIA_VISIBLE_DEVICES=all
 ENV NVIDIA_DRIVER_CAPABILITIES=compute,utility
-ENV NVIDIA_REQUIRE_CUDA="cuda>=10.0"
+ENV NVIDIA_REQUIRE_CUDA="cuda>=$CUDA_MAJOR_VERSION.$CUDA_MINOR_VERSION"
 RUN apt-get update && apt-get install -y --no-install-recommends \
       cuda-cupti-$CUDA_PKG_VERSION \
       cuda-cudart-$CUDA_PKG_VERSION \
@@ -36,11 +39,11 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
       cuda-nvml-dev-$CUDA_PKG_VERSION \
       cuda-minimal-build-$CUDA_PKG_VERSION \
       cuda-command-line-tools-$CUDA_PKG_VERSION \
-      libcudnn7=7.5.0.56-1+cuda10.0 \
-      libcudnn7-dev=7.5.0.56-1+cuda10.0 \
-      libnccl2=2.4.2-1+cuda10.0 \
-      libnccl-dev=2.4.2-1+cuda10.0 && \
-    ln -s /usr/local/cuda-10.0 /usr/local/cuda && \
+      libcudnn7=7.6.5.32-1+cuda$CUDA_MAJOR_VERSION.$CUDA_MINOR_VERSION \
+      libcudnn7-dev=7.6.5.32-1+cuda$CUDA_MAJOR_VERSION.$CUDA_MINOR_VERSION \
+      libnccl2=2.5.6-1+cuda$CUDA_MAJOR_VERSION.$CUDA_MINOR_VERSION \
+      libnccl-dev=2.5.6-1+cuda$CUDA_MAJOR_VERSION.$CUDA_MINOR_VERSION && \
+    ln -s /usr/local/cuda-$CUDA_MAJOR_VERSION.$CUDA_MINOR_VERSION /usr/local/cuda && \
     ln -s /usr/local/cuda/lib64/stubs/libcuda.so /usr/local/cuda/lib64/stubs/libcuda.so.1 && \
     /tmp/clean-layer.sh
 
@@ -67,7 +70,7 @@ RUN pip uninstall -y lightgbm && \
 
 # Install JAX
 ENV JAX_PYTHON_VERSION=cp36
-ENV JAX_CUDA_VERSION=cuda100
+ENV JAX_CUDA_VERSION=cuda$CUDA_MAJOR_VERSION$CUDA_MINOR_VERSION
 ENV JAX_PLATFORM=linux_x86_64
 ENV JAX_BASE_URL="https://storage.googleapis.com/jax-releases"
 
@@ -80,15 +83,15 @@ RUN pip uninstall -y tensorflow && \
     pip install /tmp/tensorflow_gpu/tensorflow*.whl && \
     rm -rf /tmp/tensorflow_gpu && \
     conda remove --force -y pytorch torchvision torchaudio cpuonly && \
-    conda install -y pytorch torchvision torchaudio cudatoolkit=10.0 -c pytorch && \
+    conda install -y pytorch torchvision torchaudio cudatoolkit=CUDA_MAJOR_VERSION.CUDA_MINOR_VERSION -c pytorch && \
     pip uninstall -y mxnet && \
     # b/126259508 --no-deps prevents numpy from being downgraded.
-    pip install --no-deps mxnet-cu100 && \
+    pip install --no-deps mxnet-cu$CUDA_MAJOR_VERSION$CUDA_MINOR_VERSION && \
     /tmp/clean-layer.sh
 
 # Install GPU-only packages
 RUN pip install pycuda && \
-    pip install cupy-cuda100 && \
+    pip install cupy-cuda$CUDA_MAJOR_VERSION$CUDA_MINOR_VERSION && \
     pip install pynvrtc && \
     /tmp/clean-layer.sh
 

tensorflow-whl/CHANGELOG.md

@@ -10,3 +10,4 @@
 2.1.0-rc0-py36: TensorFlow 2.1.0-rc0 with Python 3.6
 2.1.0-rc2-py36: TensorFlow 2.1.0-rc2 with Python 3.6
 2.1.0-py36: TensorFlow 2.1.0 with Python 3.6
+2.1.0-py36-2: TensorFlow 2.1.0 with CUDA 10.1