Merge pull request #16 from EntropyOrg/docker-gpu

GPU version of Docker images

Author: zmughal

.github/workflows/build-container.yml

@@ -29,6 +29,9 @@ jobs:
       uses: docker/[email protected]
       with:
         files: docker/docker-bake.hcl
+        targets: |
+          default
+          gpu
         push: ${{ github.event_name != 'pull_request' }}
         set: |
           *.cache-from=type=gha

docker/Dockerfile

@@ -1,6 +1,6 @@
 #syntax=docker/dockerfile-upstream:1.4.0-rc1
 
-ARG TF_VERSION="2.11.0"
+ARG TF_VERSION
 
 # Packages from Ubuntu focal base
 ARG APT_PKGS_PERL="perl cpanminus"
@@ -11,8 +11,6 @@ ARG APT_PKGS_DEV="libffi-dev libssl-dev pkg-config"
 
 FROM perldocker/perl-tester:5.36 AS builder
 
-ARG TF_VERSION
-
 ENV ALIEN_LIBTENSORFLOW_FROM_BINARY_VERSION=${TF_VERSION}
 
 #COPY . /build
@@ -28,8 +26,6 @@ RUN dzil build --in dist
 
 FROM tensorflow/tensorflow:${TF_VERSION}-jupyter AS base-builder
 
-ARG TF_VERSION
-
 ARG APT_PKGS_PERL
 ARG APT_PKGS_RUN
 ARG APT_PKGS_DEV
@@ -38,6 +34,9 @@ COPY --from=builder /build/dist /build/dist
 
 ENV ALIEN_LIBTENSORFLOW_FROM_BINARY_VERSION=${TF_VERSION}
 
+# Build the base with CPU
+ENV ALIEN_LIBTENSORFLOW_DEVICE_TYPE="cpu"
+
 RUN rm -Rfv /tf
 RUN mkdir -p /tf
 
@@ -48,20 +47,26 @@ RUN cpanm -nq App::plx App::cpm
 # Init plx in /tf.
 RUN bash -c 'cd /tf \
 	&& plx --init \
-	&& plx --config libspec add 10perl5-nb.ll     /perl5-nb && perl -Mlocal::lib=/perl5-nb \
+	&& plx --config libspec add 20perl5-nb.ll     /perl5-nb    && perl -Mlocal::lib=/perl5-nb \
+	&& plx --config libspec add 10perl5-libtf.ll  /perl5-libtf && perl -Mlocal::lib=/perl5-libtf \
 	&& plx --config libspec add 00perl5-global.ll /perl5 \
 	&& plx --cpanm -l /perl5 -nq App::cpm \
 	'
 
 # Install Jupyter kernel
-RUN cpm install -L /perl5 Devel::IPerl
+RUN plx cpm install -L /perl5 Devel::IPerl IO::Async::Loop::Epoll
+
+# Install libtensorflow (with Alien deps to /perl5 and actual Alien dist to
+# /perl5-libtf)
+RUN plx --cpanm -l /perl5 --installdeps --no-man-pages -nq Alien::Libtensorflow \
+    && plx --cpanm -l /perl5-libtf --no-man-pages -nq Alien::Libtensorflow
 
 # Install the actual dist.
-RUN cpm install -L /perl5 /build/dist \
+RUN plx --cpanm -l /perl5 -nq --no-man-pages /build/dist \
 	&& rm -Rf /build/dist
 
 # Clean up
-RUN apt-get remove -y $NB_APT_PKGS_DEV && apt-get autoremove
+RUN apt-get remove -y $APT_PKGS_DEV && apt-get autoremove -y
 
 RUN apt-get -qq purge \
     && apt-get -qq clean \
@@ -96,8 +101,9 @@ RUN groupadd \
         --uid ${NB_UID} \
         ${NB_USER}
 
-COPY --from=base-builder /perl5 /perl5
-COPY --from=base-builder /tf    /tf
+COPY --from=base-builder /perl5       /perl5
+COPY --from=base-builder /perl5-libtf /perl5-libtf
+COPY --from=base-builder /tf          /tf
 
 # Get rid of upstream tutorials.
 RUN rm -Rf /tf/tensorflow-tutorials

docker/Dockerfile.with-gpu-libtf

@@ -0,0 +1,75 @@
+#syntax=docker/dockerfile-upstream:1.4.0-rc1
+
+ARG TF_VERSION
+ARG TF_DEVICE_TAG
+
+FROM nb-base AS nb-builder
+
+ARG TF_VERSION
+ARG TF_DEVICE_TYPE
+
+ENV ALIEN_LIBTENSORFLOW_FROM_BINARY_VERSION=${TF_VERSION}
+ENV ALIEN_LIBTENSORFLOW_DEVICE_TYPE=${TF_DEVICE_TYPE}
+
+USER root
+
+RUN rm -Rf /perl5-libtf \
+	&& plx --cpanm -l /perl5-libtf --no-man-pages -nq Alien::Libtensorflow
+
+FROM tensorflow/tensorflow:${TF_VERSION}${TF_DEVICE_TAG}-jupyter AS nb
+
+# Packages from Ubuntu focal base
+ARG APT_PKGS_PERL="perl cpanminus"
+ARG APT_PKGS_RUN="libffi7    libssl1.1  libzmq5"
+
+ARG NB_APT_PKGS_RUN
+
+ARG NB_USER
+ARG NB_UID
+
+# Use bash as default shell, rather than sh
+ENV SHELL /bin/bash
+
+# Set up user
+ENV USER ${NB_USER}
+ENV HOME /home/${NB_USER}
+
+USER root
+
+RUN groupadd \
+        --gid ${NB_UID} \
+        ${NB_USER} && \
+    useradd \
+        --comment "Default user" \
+        --create-home \
+        --gid ${NB_UID} \
+        --no-log-init \
+        --shell /bin/bash \
+        --uid ${NB_UID} \
+        ${NB_USER}
+
+COPY --from=nb-builder /perl5       /perl5
+COPY --from=nb-builder /perl5-nb    /perl5-nb
+COPY --from=nb-builder /perl5-libtf /perl5-libtf
+COPY --from=nb-builder /tf          /tf
+
+RUN apt-get update && apt-get install -y --no-install-recommends $APT_PKGS_PERL $APT_PKGS_RUN $NB_APT_PKGS_RUN
+
+RUN cpanm -nq App::plx App::cpm
+
+RUN apt-get -qq purge \
+    && apt-get -qq clean \
+    && rm -rf /var/lib/apt/lists/*
+
+RUN rm -Rf $HOME/.perl-cpm/ $HOME/.cpanm
+
+USER ${NB_USER}
+
+RUN plx iperl --version # install kernel and exit
+
+EXPOSE 8888
+
+ENTRYPOINT [ "plx", "--base", "/tf", "--exec" ]
+CMD [ "jupyter", "notebook", "--notebook-dir=/tf", "--ip", "0.0.0.0", "--no-browser" ]
+
+# vim:ft=dockerfile

docker/docker-bake.hcl

@@ -1,12 +1,29 @@
 group "default" {
   targets = [
+    # CPU
     "base",
     "nb-omnibus",
     "nb-image-class",
     "nb-gene-expr-pred",
   ]
 }
 
+group "gpu" {
+  targets = [
+    # GPU
+    "gpu-nb-omnibus",
+  ]
+}
+
+variable "TF_VERSION" {
+  default = "2.12.0"
+}
+
+function "TF_LATEST_VERSION" {
+  params = []
+  result = "2.12.0" # current upstream version
+}
+
 variable "REGISTRY" {
   default = "quay.io/entropyorg/"
 }
@@ -15,10 +32,6 @@ variable "IMAGE" {
   default = "perl-ai-tensorflow-libtensorflow"
 }
 
-variable "TAG" {
-  default = "latest"
-}
-
 variable "NB_USER" {
   default = "jovyan"
 }
@@ -27,47 +40,95 @@ variable "NB_UID" {
   default = 1000
 }
 
-target "nb" {
+function "device2tag" {
+  params = [device]
+  result = equal(device,"cpu") ? "" : equal(device,"gpu") ? "-gpu" : ""
+}
+
+function "tag" {
+  params = [device, suffix]
+  result = [
+    "${REGISTRY}${IMAGE}:${TF_VERSION}${device2tag(device)}${suffix}",
+    equal(TF_VERSION, TF_LATEST_VERSION())
+      ? "${REGISTRY}${IMAGE}:latest${device2tag(device)}${suffix}"
+      : "",
+  ]
+}
+
+target "nb-all" {
   args = {
     NB_USER = NB_USER
     NB_UID  = NB_UID
+    TF_VERSION = TF_VERSION
+  }
+}
+
+target "nb-cpu" {
+  inherits   = [ "nb-all" ]
+  args = {
+    TF_DEVICE_TYPE = "cpu"
+    TF_DEVICE_TAG  = "${device2tag("cpu")}"
+  }
+}
+
+target "nb-gpu" {
+  inherits   = [ "nb-all" ]
+  args = {
+    TF_DEVICE_TYPE = "gpu"
+    TF_DEVICE_TAG  = "${device2tag("gpu")}"
   }
 }
 
 target "base" {
-  inherits   = [ "nb" ]
+  inherits   = [ "nb-cpu" ]
   dockerfile = "docker/Dockerfile"
   target     = "base"
-  tags = ["${REGISTRY}${IMAGE}:${TAG}"]
+  tags = tag("cpu", "")
 }
 
 target "nb-image-class" {
-  inherits   = [ "nb" ]
+  inherits   = [ "nb-cpu" ]
   dockerfile = "docker/Dockerfile.nb-image-class"
   target     = "image-classification"
   contexts = {
     base           = "target:base"
   }
-  tags = ["${REGISTRY}${IMAGE}:${TAG}-nb-image-class"]
+  tags = tag("cpu", "-nb-image-class")
 }
 
 target "nb-gene-expr-pred" {
-  inherits   = [ "nb" ]
+  inherits   = [ "nb-cpu" ]
   dockerfile = "docker/Dockerfile.nb-gene-expr-pred"
   target     = "gene-expression-prediction"
   contexts = {
     base           = "target:base"
   }
-  tags = ["${REGISTRY}${IMAGE}:${TAG}-nb-gene-expr-pred"]
+  tags = tag("cpu", "-nb-gene-expr-pred")
 }
 
 target "nb-omnibus" {
-  inherits   = [ "nb" ]
+  inherits   = [ "nb-cpu" ]
   dockerfile = "docker/Dockerfile.nb-omnibus"
   contexts = {
     base               = "target:base"
     nb-image-class     = "target:nb-image-class"
     nb-gene-expr-pred  = "target:nb-gene-expr-pred"
   }
-  tags = ["${REGISTRY}${IMAGE}:${TAG}-nb-omnibus"]
+  tags = tag("cpu", "-nb-omnibus")
+}
+
+target "gpu-nb-omnibus" {
+  inherits   = [ "nb-gpu" ]
+  args = {
+    NB_APT_PKGS_RUN = <<EOF
+      libjpeg-turbo8     libpng16-16 libgsl23
+      samtools libhts3 tabix libxml2 libexpat1 libdb5.3 netpbm gnuplot-nox
+    EOF
+  }
+  dockerfile = "docker/Dockerfile.with-gpu-libtf"
+  target     = "nb"
+  contexts = {
+    nb-base            = "target:nb-omnibus"
+  }
+  tags = tag("gpu", "-nb-omnibus")
 }