Update frontpage.html

[skip CI]

.github/actions/setup_env_vars/action.yml

@@ -0,0 +1,57 @@
+name: "Setup Env Vars"
+
+runs:
+  using: "composite"
+  steps:
+    - name: Get Commit SHA (For Push Event)
+      if: ${{ github.event_name == 'push' }}
+      shell: bash
+      env:
+        SHA: ${{ github.sha }}
+        git_repo_full: ${{ github.repository }}
+        pr_number: ""
+      run: |
+        short_sha=$(git rev-parse --short "$SHA")
+        echo "SHORT_SHA=$short_sha" >> $GITHUB_ENV
+        echo "PR_NUMBER=$pr_number" >> $GITHUB_ENV
+
+        target_branch=${GITHUB_REF##*/}
+        echo "TARGET_BRANCH=$target_branch" >> $GITHUB_ENV
+
+        repo_name=${git_repo_full##*/}
+        echo "REPO_NAME=$repo_name" >> $GITHUB_ENV
+
+        IFS='-' read -ra name_parts <<< "$repo_name"
+        echo "LANG=${name_parts[1]}" >> $GITHUB_ENV
+
+        job_name=${repo_name}/${target_branch}
+        echo "JOB_NAME=$job_name" >> $GITHUB_ENV
+
+        echo "CACHE_DIR=ci_cache_push" >> $GITHUB_ENV
+
+
+    - name: Get Commit SHA (For Pull Request)
+      if: ${{ github.event_name == 'pull_request_target' }}
+      shell: bash
+      env:
+        SHA: ${{ github.event.pull_request.head.sha }}
+        target_branch: ${{ github.event.pull_request.base.ref }}
+        git_repo_full: ${{ github.event.pull_request.base.repo.full_name }}
+        pr_number: PR-${{ github.event.number }}
+      run: |
+        short_sha=$(git rev-parse --short "$SHA")
+        echo "SHORT_SHA=$short_sha" >> $GITHUB_ENV
+        echo "PR_NUMBER=$pr_number" >> $GITHUB_ENV
+
+        echo "TARGET_BRANCH=$target_branch" >> $GITHUB_ENV
+        
+        repo_name=${git_repo_full##*/}
+        echo "REPO_NAME=$repo_name" >> $GITHUB_ENV
+
+        IFS='-' read -ra name_parts <<< "$repo_name"
+        echo "LANG=${name_parts[1]}" >> $GITHUB_ENV
+
+        job_name=${repo_name}/${pr_number}/${short_sha}
+        echo "JOB_NAME=$job_name" >> $GITHUB_ENV
+
+        echo "CACHE_DIR=ci_cache_pr" >> $GITHUB_ENV

.github/actions/submit-job/action.yml

@@ -0,0 +1,91 @@
+name: "Submit Job to AWS Batch"
+inputs:
+  job-type:
+    required: true
+  job-name:
+    required: true
+  work-dir:
+    required: false
+    default: .
+  command:
+    required: true
+
+runs:
+  using: "composite"
+  steps:
+    - name: Configure AWS Credentials
+      uses: aws-actions/configure-aws-credentials@v2
+      with:
+        role-to-assume: arn:aws:iam::650140442593:role/D2L_CI_Batch
+        role-duration-seconds: 14400  # this requires changing max session duration to 4hrs in AWS Console for D2L_CI_Batch
+        aws-region: us-west-2
+
+    - name: Install dependencies
+      shell: bash
+      run: |
+        pip install boto3
+
+    - name: Check for Actor Permissions
+      id: check
+      continue-on-error: true
+      uses: prince-chrismc/check-actor-permissions-action@88995341900ae4e7aa248df51d32aeb981cc62f5 # v2
+      with:
+        github_token: ${{ github.token }}
+        permission: write
+
+    - name: Submit Job (For Push on development branches)
+      if: ${{ github.event_name == 'push' && github.ref != 'refs/heads/release' && github.ref != 'refs/heads/classic' }}
+      shell: bash
+      run: |
+        echo "Start submitting job for a Push Event on a Development Branch"
+        # Add "-push" for all these jobs to use elevated push level job-def permissions
+        python ./ci/submit-job.py --job-type ${{ inputs.job-type }}-push \
+                                  --name ${{ inputs.job-name }}-'${{ github.ref }}' \
+                                  --source-ref '${{ github.ref }}' \
+                                  --work-dir ${{ inputs.work-dir }} \
+                                  --remote https://github.com/'${{ github.repository }}' \
+                                  --command "${{ inputs.command }}" \
+                                  --safe-to-use-script \
+                                  --wait
+
+    - name: Submit Job (For Push on Release/Classic)
+      if: ${{ github.event_name == 'push' && (github.ref == 'refs/heads/release' || github.ref == 'refs/heads/classic') }}
+      shell: bash
+      run: |
+        echo "Start submitting job for a Push Event on Release/Classic Branch"
+        # Add "-release" for all these jobs to use elevated release level job-def permissions
+        python ./ci/submit-job.py --job-type ${{ inputs.job-type }}-release \
+                                  --name ${{ inputs.job-name }}-'${{ github.ref }}' \
+                                  --source-ref '${{ github.ref }}' \
+                                  --work-dir ${{ inputs.work-dir }} \
+                                  --remote https://github.com/'${{ github.repository }}' \
+                                  --command "${{ inputs.command }}" \
+                                  --safe-to-use-script \
+                                  --wait
+
+    - name: Submit Job (For Pull Request Safe Scripts)
+      if: ${{ github.event_name == 'pull_request_target' && steps.check.outputs.permitted == 'true' }}
+      shell: bash
+      run: |
+        echo "Start submitting job for PR Safe Scripts"
+        python ./ci/submit-job.py --job-type ${{ inputs.job-type }} \
+                                  --name ${{ inputs.job-name }}-PR#'${{ github.event.number }}' \
+                                  --source-ref '${{ github.event.pull_request.head.sha }}' \
+                                  --work-dir ${{ inputs.work-dir }} \
+                                  --remote https://github.com/'${{ github.event.pull_request.head.repo.full_name }}' \
+                                  --command "${{ inputs.command }}" \
+                                  --safe-to-use-script \
+                                  --wait
+
+    - name: Submit Job (For Pull Request Not Safe Scripts)
+      if: ${{ github.event_name == 'pull_request_target' && steps.check.outputs.permitted != 'true' }}
+      shell: bash
+      run: |
+        echo "Start submitting job for PR Unsafe Scripts"
+        python ./ci/submit-job.py --job-type ${{ inputs.job-type }} \
+                                  --name ${{ inputs.job-name }}-PR#'${{ github.event.number }}' \
+                                  --source-ref '${{ github.event.pull_request.head.sha }}' \
+                                  --work-dir ${{ inputs.work-dir }} \
+                                  --remote https://github.com/'${{ github.event.pull_request.head.repo.full_name }}' \
+                                  --command "${{ inputs.command }}" \
+                                  --wait

.github/workflow_scripts/build_and_deploy.sh

@@ -0,0 +1,45 @@
+#!/bin/bash
+
+# Move all scripts related to html here!
+
+set -ex
+
+REPO_NAME="$1"  # Eg. 'd2l-zh'
+TARGET_BRANCH="$2" # Eg. 'master' ; if PR raised to master
+JOB_NAME="$3" # Eg. 'd2l-zh/master' or 'd2l-zh/PR-2453/21be1a4'
+LANG="$4" # Eg. 'en','zh' etc.
+CACHE_DIR="$5"  # Eg. 'ci_cache_pr' or 'ci_cache_push'
+
+pip3 install d2l==0.17.6
+mkdir _build
+
+source $(dirname "$0")/utils.sh
+
+# Move aws copy commands for cache restore outside
+measure_command_time "aws s3 sync s3://preview.d2l.ai/"$CACHE_DIR"/"$REPO_NAME"-"$TARGET_BRANCH"/_build _build --delete --quiet --exclude 'eval*/data/*'"
+
+# Build D2L Website
+./.github/workflow_scripts/build_html.sh
+
+# Build PDFs
+d2lbook build pdf
+d2lbook build pdf --tab pytorch
+
+
+# Check if the JOB_NAME is "$REPO_NAME/release"
+if [[ "$JOB_NAME" == "$REPO_NAME/release" ]]; then
+
+  # Setup D2L Bot
+  source $(dirname "$0")/setup_git.sh
+  setup_git
+
+  d2lbook build pkg
+  d2lbook deploy html pdf pkg colab sagemaker slides --s3 "s3://${LANG}-v2.d2l.ai/"
+
+else
+  # Run d2lbook preview deployment
+  d2lbook deploy html pdf --s3 "s3://preview.d2l.ai/${JOB_NAME}/"
+fi
+
+# Move aws copy commands for cache store outside
+measure_command_time "aws s3 sync _build s3://preview.d2l.ai/"$CACHE_DIR"/"$REPO_NAME"-"$TARGET_BRANCH"/_build --acl public-read --quiet --exclude 'eval*/data/*'"

.github/workflow_scripts/build_mxnet.sh

@@ -0,0 +1,44 @@
+#!/bin/bash
+
+set -ex
+
+# Used to capture status exit of build eval command
+ss=0
+
+REPO_NAME="$1"  # Eg. 'd2l-zh'
+TARGET_BRANCH="$2" # Eg. 'master' ; if PR raised to master
+CACHE_DIR="$3"  # Eg. 'ci_cache_pr' or 'ci_cache_push'
+
+pip3 install d2l==0.17.6
+mkdir _build
+
+source $(dirname "$0")/utils.sh
+
+# Move sanity check outside
+d2lbook build outputcheck tabcheck
+
+# Move aws copy commands for cache restore outside
+if [ "$DISABLE_CACHE" = "false" ]; then
+  echo "Retrieving mxnet build cache from "$CACHE_DIR""
+  measure_command_time "aws s3 sync s3://preview.d2l.ai/"$CACHE_DIR"/"$REPO_NAME"-"$TARGET_BRANCH"/_build/eval/ _build/eval/ --delete --quiet --exclude 'data/*'"
+fi
+
+# MXNet training for the following notebooks is slow in the container;
+# Setting NTHREADS=4 below seems to fix the issue:
+# 1. chapter_multilayer-perceptrons/dropout.md
+# 2. chapter_multilayer-perceptrons/mlp-implementation.md
+# 3. chapter_linear-classification/softmax-regression-concise.md
+# 4. chapter_linear-classification/softmax-regression-scratch.md
+export MXNET_CPU_WORKER_NTHREADS=4
+# Continue the script even if some notebooks in build fail to
+# make sure that cache is copied to s3 for the successful notebooks
+d2lbook build eval || ((ss=1))
+
+# Move aws copy commands for cache store outside
+echo "Upload mxnet build cache to s3"
+measure_command_time "aws s3 sync _build s3://preview.d2l.ai/"$CACHE_DIR"/"$REPO_NAME"-"$TARGET_BRANCH"/_build --acl public-read --quiet --exclude 'eval*/data/*'"
+
+# Exit with a non-zero status if evaluation failed
+if [ "$ss" -ne 0 ]; then
+  exit 1
+fi

.github/workflow_scripts/build_paddle.sh

@@ -0,0 +1,37 @@
+#!/bin/bash
+
+set -ex
+
+# Used to capture status exit of build eval command
+ss=0
+
+REPO_NAME="$1"  # Eg. 'd2l-zh'
+TARGET_BRANCH="$2" # Eg. 'master' ; if PR raised to master
+CACHE_DIR="$3"  # Eg. 'ci_cache_pr' or 'ci_cache_push'
+
+pip3 install d2l==0.17.6
+mkdir _build
+
+source $(dirname "$0")/utils.sh
+
+# Move sanity check outside
+d2lbook build outputcheck tabcheck
+
+# Move aws copy commands for cache restore outside
+if [ "$DISABLE_CACHE" = "false" ]; then
+  echo "Retrieving paddle build cache from "$CACHE_DIR""
+  measure_command_time "aws s3 sync s3://preview.d2l.ai/"$CACHE_DIR"/"$REPO_NAME"-"$TARGET_BRANCH"/_build/eval_paddle/ _build/eval_paddle/ --delete --quiet --exclude 'data/*'"
+fi
+
+# Continue the script even if some notebooks in build fail to
+# make sure that cache is copied to s3 for the successful notebooks
+d2lbook build eval --tab paddle || ((ss=1))
+
+# Move aws copy commands for cache store outside
+echo "Upload paddle build cache to s3"
+measure_command_time "aws s3 sync _build s3://preview.d2l.ai/"$CACHE_DIR"/"$REPO_NAME"-"$TARGET_BRANCH"/_build --acl public-read --quiet --exclude 'eval*/data/*'"
+
+# Exit with a non-zero status if evaluation failed
+if [ "$ss" -ne 0 ]; then
+  exit 1
+fi

.github/workflow_scripts/build_pytorch.sh

@@ -0,0 +1,41 @@
+#!/bin/bash
+
+set -ex
+
+# Used to capture status exit of build eval command
+ss=0
+
+REPO_NAME="$1"  # Eg. 'd2l-zh'
+TARGET_BRANCH="$2" # Eg. 'master' ; if PR raised to master
+CACHE_DIR="$3"  # Eg. 'ci_cache_pr' or 'ci_cache_push'
+
+pip3 install d2l==0.17.6
+pip3 install numpy==1.24.4  # numpy version installed by d2l==0.17.6 is too old
+mkdir _build
+
+source $(dirname "$0")/utils.sh
+
+# Move sanity check outside
+d2lbook build outputcheck tabcheck
+
+# Move aws copy commands for cache restore outside
+if [ "$DISABLE_CACHE" = "false" ]; then
+  echo "Retrieving pytorch build cache from "$CACHE_DIR""
+  measure_command_time "aws s3 sync s3://preview.d2l.ai/"$CACHE_DIR"/"$REPO_NAME"-"$TARGET_BRANCH"/_build/eval_pytorch/ _build/eval_pytorch/ --delete --quiet --exclude 'data/*'"
+  echo "Retrieving pytorch slides cache from "$CACHE_DIR""
+  aws s3 sync s3://preview.d2l.ai/"$CACHE_DIR"/"$REPO_NAME"-"$TARGET_BRANCH"/_build/slides _build/slides --delete --quiet --exclude 'data/*'
+fi
+
+# Continue the script even if some notebooks in build fail to
+# make sure that cache is copied to s3 for the successful notebooks
+d2lbook build eval --tab pytorch || ((ss=1))
+d2lbook build slides --tab pytorch
+
+# Move aws copy commands for cache store outside
+echo "Upload pytorch build cache to s3"
+measure_command_time "aws s3 sync _build s3://preview.d2l.ai/"$CACHE_DIR"/"$REPO_NAME"-"$TARGET_BRANCH"/_build --acl public-read --quiet --exclude 'eval*/data/*'"
+
+# Exit with a non-zero status if evaluation failed
+if [ "$ss" -ne 0 ]; then
+  exit 1
+fi

.github/workflow_scripts/build_tf.sh

@@ -0,0 +1,39 @@
+#!/bin/bash
+
+set -ex
+
+# Used to capture status exit of build eval command
+ss=0
+
+REPO_NAME="$1"  # Eg. 'd2l-zh'
+TARGET_BRANCH="$2" # Eg. 'master' ; if PR raised to master
+CACHE_DIR="$3"  # Eg. 'ci_cache_pr' or 'ci_cache_push'
+
+pip3 install d2l==0.17.6
+mkdir _build
+
+source $(dirname "$0")/utils.sh
+
+# Move sanity check outside
+d2lbook build outputcheck tabcheck
+
+# Move aws copy commands for cache restore outside
+if [ "$DISABLE_CACHE" = "false" ]; then
+  echo "Retrieving tensorflow build cache from "$CACHE_DIR""
+  measure_command_time "aws s3 sync s3://preview.d2l.ai/"$CACHE_DIR"/"$REPO_NAME"-"$TARGET_BRANCH"/_build/eval_tensorflow/ _build/eval_tensorflow/ --delete --quiet --exclude 'data/*'"
+fi
+
+export TF_CPP_MIN_LOG_LEVEL=3
+export TF_FORCE_GPU_ALLOW_GROWTH=true
+# Continue the script even if some notebooks in build fail to
+# make sure that cache is copied to s3 for the successful notebooks
+d2lbook build eval --tab tensorflow || ((ss=1))
+
+# Move aws copy commands for cache store outside
+echo "Upload tensorflow build cache to s3"
+measure_command_time "aws s3 sync _build s3://preview.d2l.ai/"$CACHE_DIR"/"$REPO_NAME"-"$TARGET_BRANCH"/_build --acl public-read --quiet --exclude 'eval*/data/*'"
+
+# Exit with a non-zero status if evaluation failed
+if [ "$ss" -ne 0 ]; then
+  exit 1
+fi

.github/workflow_scripts/setup_git.sh

@@ -0,0 +1,18 @@
+function setup_git {
+    # Turn off logging
+    set +x
+    mkdir -p $HOME/.ssh
+    echo "yes" | ssh-keyscan -t rsa github.com >> ~/.ssh/known_hosts
+
+    # Retrieve the SSH key securely from AWS Secrets Manager
+    GIT_SSH_KEY=$(aws secretsmanager get-secret-value --secret-id d2l_bot_github --query SecretString --output text --region us-west-2)
+
+    # Write the SSH key to a file
+    echo "$GIT_SSH_KEY" > $HOME/.ssh/id_rsa
+    chmod 600 $HOME/.ssh/id_rsa
+
+    git config --global user.name "d2l-bot"
+    git config --global user.email "100248899+d2l-bot@users.noreply.github.com"
+
+    echo "Successfully Configured Bot"
+}

.github/workflow_scripts/utils.sh

@@ -0,0 +1,26 @@
+#!/bin/bash
+
+# By default, all builds are cached
+DISABLE_CACHE=false  # Eg. 'true' or 'false'
+
+
+# Function to measure command execution time
+measure_command_time() {
+    local command="$1"
+
+    # Start timing
+    local start_time=$(date +%s)
+
+    # Run the command
+    eval "$command"
+
+    # Calculate the time taken
+    local end_time=$(date +%s)
+    local elapsed_time=$((end_time - start_time))
+
+    # Format the elapsed time for display
+    local formatted_time=$(printf "%02dhr %02dmin %02dsec" $((elapsed_time / 3600)) $(((elapsed_time % 3600) / 60)) $((elapsed_time % 60)))
+
+    # Print the elapsed time
+    echo "Time taken for $command: $formatted_time"
+}

.github/workflows/build-docker.yml

@@ -0,0 +1,88 @@
+name: Build Docker Image
+
+on:
+  workflow_dispatch:
+    inputs:
+      image_torch:
+        type: boolean
+        description: Build PyTorch Image
+      image_tf:
+        type: boolean
+        description: Build TensorFlow Image
+      image_mxnet:
+        type: boolean
+        description: Build MXNet Image
+      image_paddle:
+        type: boolean
+        description: Build Paddle Image
+      image_builder:
+        type: boolean
+        description: Build D2L Builder Image
+
+permissions:
+  id-token: write
+  contents: read
+
+jobs:
+  build_docker_image:
+    name: Build D2L Docker Images
+    runs-on: ubuntu-latest
+    defaults:
+      run:
+        working-directory: ./ci/docker
+    steps:
+    - uses: actions/checkout@v3
+    - name: Configure AWS Credentials
+      uses: aws-actions/configure-aws-credentials@v2
+      with:
+        role-to-assume: arn:aws:iam::650140442593:role/D2L_CI_Docker
+        role-duration-seconds: 3600
+        aws-region: us-west-2
+
+    - name: Build D2L PyTorch Image
+      if: github.event.inputs.image_torch == 'true'
+      run: |
+        chmod +x ./login_ecr.sh; ./login_ecr.sh
+        docker build -f Dockerfile.d2l-zh-torch -t d2l-containers:d2l-zh-torch-latest .
+        docker tag d2l-containers:d2l-zh-torch-latest 650140442593.dkr.ecr.us-west-2.amazonaws.com/d2l-containers:d2l-zh-torch-latest
+        docker push 650140442593.dkr.ecr.us-west-2.amazonaws.com/d2l-containers:d2l-zh-torch-latest
+        # Clean up to reclaim space
+        echo "y" | docker system prune -a
+
+    - name: Build D2L TensorFlow Image
+      if: github.event.inputs.image_tf == 'true'
+      run: |
+        chmod +x ./login_ecr.sh; ./login_ecr.sh
+        docker build -f Dockerfile.d2l-zh-tf -t d2l-containers:d2l-zh-tensorflow-latest .
+        docker tag d2l-containers:d2l-zh-tensorflow-latest 650140442593.dkr.ecr.us-west-2.amazonaws.com/d2l-containers:d2l-zh-tensorflow-latest
+        docker push 650140442593.dkr.ecr.us-west-2.amazonaws.com/d2l-containers:d2l-zh-tensorflow-latest
+        # Clean up to reclaim space
+        echo "y" | docker system prune -a
+
+    - name: Build D2L MXNet Image
+      if: github.event.inputs.image_mxnet == 'true'
+      run: |
+        chmod +x ./login_ecr.sh; ./login_ecr.sh
+        docker build -f Dockerfile.d2l-zh-mxnet -t d2l-containers:d2l-zh-mxnet-latest .
+        docker tag d2l-containers:d2l-zh-mxnet-latest 650140442593.dkr.ecr.us-west-2.amazonaws.com/d2l-containers:d2l-zh-mxnet-latest
+        docker push 650140442593.dkr.ecr.us-west-2.amazonaws.com/d2l-containers:d2l-zh-mxnet-latest
+        # Clean up to reclaim space
+        echo "y" | docker system prune -a
+
+    - name: Build D2L Paddle Image
+      if: github.event.inputs.image_paddle == 'true'
+      run: |
+        chmod +x ./login_ecr.sh; ./login_ecr.sh
+        docker build -f Dockerfile.d2l-zh-paddle -t d2l-containers:d2l-zh-paddle-latest .
+        docker tag d2l-containers:d2l-zh-paddle-latest 650140442593.dkr.ecr.us-west-2.amazonaws.com/d2l-containers:d2l-zh-paddle-latest
+        docker push 650140442593.dkr.ecr.us-west-2.amazonaws.com/d2l-containers:d2l-zh-paddle-latest
+        # Clean up to reclaim space
+        echo "y" | docker system prune -a
+
+    - name: Build D2L CPU Builder Image
+      if: github.event.inputs.image_builder == 'true'
+      run: |
+        chmod +x ./login_ecr.sh; ./login_ecr.sh
+        docker build -f Dockerfile.d2l-builder -t d2l-containers:d2l-builder-latest .
+        docker tag d2l-containers:d2l-builder-latest 650140442593.dkr.ecr.us-west-2.amazonaws.com/d2l-containers:d2l-builder-latest
+        docker push 650140442593.dkr.ecr.us-west-2.amazonaws.com/d2l-containers:d2l-builder-latest

.github/workflows/ci.yml

@@ -0,0 +1,142 @@
+name: Continuous Integration
+
+on:
+  # Triggers the workflow on push or pull request events only for the specified branches
+  push:
+    branches:
+      - master
+      - release
+  pull_request_target:
+    branches:
+      - master
+      - release
+
+permissions:
+  id-token: write
+  pull-requests: write
+
+defaults:
+  run:
+    shell: bash
+
+jobs:
+  build_torch:
+      name: Build PyTorch
+      if: "github.repository == 'd2l-ai/d2l-zh' && !contains(github.event.head_commit.message, '[skip torch]') && !contains(github.event.head_commit.message, '[skip frameworks]')"
+      runs-on: ubuntu-latest
+      steps:
+        - name: Checkout repository
+          uses: actions/checkout@v3
+        - name: Setup Env Vars
+          uses: ./.github/actions/setup_env_vars
+        - name: Evaluate PyTorch on AWS Batch
+          uses: ./.github/actions/submit-job
+          with:
+            job-type: ci-gpu-torch
+            job-name: D2L-Build-PyTorch
+            command: chmod +x ./.github/workflow_scripts/build_pytorch.sh && ./.github/workflow_scripts/build_pytorch.sh "${{ env.REPO_NAME }}" "${{ env.TARGET_BRANCH }}" "${{ env.CACHE_DIR }}"
+        - name: Terminate Batch Job on Cancellation
+          if: ${{ cancelled() && env.Batch_JobID }}
+          run: |
+            echo "Terminating Submitted AWS Batch Job: "${{ env.Batch_JobID }}""
+            aws batch terminate-job --job-id "${{ env.Batch_JobID }}" --reason "Job terminated by cancelled workflow"
+
+  build_tf:
+      name: Build Tensorflow
+      if: "github.repository == 'd2l-ai/d2l-zh' && !contains(github.event.head_commit.message, '[skip tf]') && !contains(github.event.head_commit.message, '[skip frameworks]')"
+      runs-on: ubuntu-latest
+      steps:
+        - name: Checkout repository
+          uses: actions/checkout@v3
+        - name: Setup Env Vars
+          uses: ./.github/actions/setup_env_vars
+        - name: Evaluate Tensorflow on AWS Batch
+          uses: ./.github/actions/submit-job
+          with:
+            job-type: ci-gpu-tf
+            job-name: D2L-Build-Tensorflow
+            command: chmod +x ./.github/workflow_scripts/build_tf.sh && ./.github/workflow_scripts/build_tf.sh "${{ env.REPO_NAME }}" "${{ env.TARGET_BRANCH }}" "${{ env.CACHE_DIR }}"
+        - name: Terminate Batch Job on Cancellation
+          if: ${{ cancelled() && env.Batch_JobID }}
+          run: |
+            echo "Terminating Submitted AWS Batch Job: "${{ env.Batch_JobID }}""
+            aws batch terminate-job --job-id "${{ env.Batch_JobID }}" --reason "Job terminated by cancelled workflow"
+
+  build_mxnet:
+      name: Build MXNet
+      if: "github.repository == 'd2l-ai/d2l-zh' && !contains(github.event.head_commit.message, '[skip mxnet]') && !contains(github.event.head_commit.message, '[skip frameworks]')"
+      runs-on: ubuntu-latest
+      steps:
+        - name: Checkout repository
+          uses: actions/checkout@v3
+        - name: Setup Env Vars
+          uses: ./.github/actions/setup_env_vars
+        - name: Evaluate MXNet on AWS Batch
+          uses: ./.github/actions/submit-job
+          with:
+            job-type: ci-gpu-mxnet
+            job-name: D2L-Build-MXNet
+            command: chmod +x ./.github/workflow_scripts/build_mxnet.sh && ./.github/workflow_scripts/build_mxnet.sh "${{ env.REPO_NAME }}" "${{ env.TARGET_BRANCH }}" "${{ env.CACHE_DIR }}"
+        - name: Terminate Batch Job on Cancellation
+          if: ${{ cancelled() && env.Batch_JobID }}
+          run: |
+            echo "Terminating Submitted AWS Batch Job: "${{ env.Batch_JobID }}""
+            aws batch terminate-job --job-id "${{ env.Batch_JobID }}" --reason "Job terminated by cancelled workflow"
+
+  build_paddle:
+      name: Build Paddle
+      if: "github.repository == 'd2l-ai/d2l-zh' && !contains(github.event.head_commit.message, '[skip paddle]') && !contains(github.event.head_commit.message, '[skip frameworks]')"
+      runs-on: ubuntu-latest
+      steps:
+        - name: Checkout repository
+          uses: actions/checkout@v3
+        - name: Setup Env Vars
+          uses: ./.github/actions/setup_env_vars
+        - name: Evaluate Paddle on AWS Batch
+          uses: ./.github/actions/submit-job
+          with:
+            job-type: ci-gpu-paddle
+            job-name: D2L-Build-Paddle
+            command: chmod +x ./.github/workflow_scripts/build_paddle.sh && ./.github/workflow_scripts/build_paddle.sh "${{ env.REPO_NAME }}" "${{ env.TARGET_BRANCH }}" "${{ env.CACHE_DIR }}"
+        - name: Terminate Batch Job on Cancellation
+          if: ${{ cancelled() && env.Batch_JobID }}
+          run: |
+            echo "Terminating Submitted AWS Batch Job: "${{ env.Batch_JobID }}""
+            aws batch terminate-job --job-id "${{ env.Batch_JobID }}" --reason "Job terminated by cancelled workflow"
+
+  build_and_deploy:
+      name: Build Website/PDF & Publish
+      needs: [build_torch, build_tf, build_mxnet, build_paddle]
+      if: |
+        always() &&
+        github.repository == 'd2l-ai/d2l-zh' &&
+        !contains(github.event.head_commit.message, '[skip builder]') &&
+        (needs.build_torch.result == 'success' || needs.build_torch.result == 'skipped') &&
+        (needs.build_tf.result == 'success' || needs.build_tf.result == 'skipped') &&
+        (needs.build_mxnet.result == 'success' || needs.build_mxnet.result == 'skipped') &&
+        (needs.build_paddle.result == 'success' || needs.build_paddle.result == 'skipped')
+      runs-on: ubuntu-latest
+      steps:
+        - name: Checkout repository
+          uses: actions/checkout@v3
+        - name: Setup Env Vars
+          uses: ./.github/actions/setup_env_vars
+        - name: Build Website & PDFs on AWS Batch
+          uses: ./.github/actions/submit-job
+          with:
+            job-type: ci-cpu
+            job-name: D2L-Builder
+            command: chmod +x ./.github/workflow_scripts/build_and_deploy.sh ./.github/workflow_scripts/build_html.sh && ./.github/workflow_scripts/build_and_deploy.sh "${{ env.REPO_NAME }}" "${{ env.TARGET_BRANCH }}" "${{ env.JOB_NAME }}" "${{ env.LANG }}" "${{ env.CACHE_DIR }}"
+        - name: Terminate Batch Job on Cancellation
+          if: ${{ cancelled() && env.Batch_JobID }}
+          run: |
+            echo "Terminating Submitted AWS Batch Job: "${{ env.Batch_JobID }}""
+            aws batch terminate-job --job-id "${{ env.Batch_JobID }}" --reason "Job terminated by cancelled workflow"
+        - name: Comment on PR
+          if: ${{ github.event_name == 'pull_request_target' }}
+          uses: peter-evans/create-or-update-comment@38e799a33166c9a254f2e3660d4d49ecd67eb80c # v3
+          with:
+            issue-number: ${{ github.event.number }}
+            body: |
+              Job PR-${{ github.event.number }}-${{ env.SHORT_SHA }} is done.
+              Check the results at http://preview.d2l.ai/${{ env.JOB_NAME }}

.github/workflows/clear-cache.yml

@@ -0,0 +1,39 @@
+name: Clear CI Cache
+
+on:
+  workflow_dispatch:
+    inputs:
+      target_branch:
+        type: choice
+        description: Choose the cache associated branch
+        options:
+        - master
+        - release
+
+      cache_dir:
+        type: choice
+        description: Choose cache type, pull requests or push events 
+        options:
+        - ci_cache_pr
+        - ci_cache_push
+
+permissions:
+  id-token: write
+  contents: read
+
+jobs:
+  clear_cache:
+    name: Clear CI Cache
+    runs-on: ubuntu-latest
+    steps:
+    - name: Configure AWS Credentials
+      uses: aws-actions/configure-aws-credentials@v2
+      with:
+        role-to-assume: arn:aws:iam::650140442593:role/D2L_CI_Clear_Cache
+        role-duration-seconds: 3600
+        aws-region: us-west-2
+    - name: Delete Cache
+      run: |
+        CACHE_S3_URL="s3://preview.d2l.ai/${{ github.event.inputs.cache_dir }}/d2l-zh-${{ github.event.inputs.target_branch }}/"
+        echo "Removing cache in $CACHE_S3_URL"
+        aws s3 rm --recursive $CACHE_S3_URL

Jenkinsfile

@@ -1,94 +0,0 @@
-stage("Build and Publish") {
-  // such as d2l-en and d2l-zh
-  def REPO_NAME = env.JOB_NAME.split('/')[0]
-  // such as en and zh
-  def LANG = REPO_NAME.split('-')[1]
-  // The current branch or the branch this PR will merge into
-  def TARGET_BRANCH = env.CHANGE_TARGET ? env.CHANGE_TARGET : env.BRANCH_NAME
-  // such as d2l-en-master
-  def TASK = REPO_NAME + '-' + TARGET_BRANCH
-  node('d2l-worker') {
-    ws("workspace/${TASK}") {
-      checkout scm
-      // conda environment
-      def ENV_NAME = "${TASK}-${EXECUTOR_NUMBER}";
-
-      sh label: "Build Environment", script: """set -ex
-      conda env update -n ${ENV_NAME} -f static/build.yml
-      conda activate ${ENV_NAME}
-      pip uninstall -y d2lbook
-      pip install git+https://github.com/d2l-ai/d2l-book
-      pip list
-      nvidia-smi
-      """
-
-      sh label: "Sanity Check", script: """set -ex
-      conda activate ${ENV_NAME}
-      d2lbook build outputcheck tabcheck
-      """
-
-      sh label: "Execute Notebooks", script: """set -ex
-      conda activate ${ENV_NAME}
-      ./static/cache.sh restore _build/eval/data
-      d2lbook build eval
-      ./static/cache.sh store _build/eval/data
-      """
-
-      sh label: "Execute Notebooks [Pytorch]", script: """set -ex
-      conda activate ${ENV_NAME}
-      ./static/cache.sh restore _build/eval_pytorch/data
-      d2lbook build eval --tab pytorch
-      d2lbook build slides --tab pytorch
-      ./static/cache.sh store _build/eval_pytorch/data
-      """
-
-      sh label: "Execute Notebooks [Tensorflow]", script: """set -ex
-      conda activate ${ENV_NAME}
-      ./static/cache.sh restore _build/eval_tensorflow/data
-      export TF_CPP_MIN_LOG_LEVEL=3
-      export TF_FORCE_GPU_ALLOW_GROWTH=true
-      d2lbook build eval --tab tensorflow
-      ./static/cache.sh store _build/eval_tensorflow/data
-      """
-
-      sh label: "Execute Notebooks [Paddlepaddle]", script: """set -ex
-      conda activate ${ENV_NAME}
-      ./static/cache.sh restore _build/eval_paddle/data
-      d2lbook build eval --tab paddle
-      ./static/cache.sh store _build/eval_paddle/data
-      """
-
-      sh label:"Build HTML", script:"""set -ex
-      conda activate ${ENV_NAME}
-      ./static/build_html.sh
-      """
-
-      sh label:"Build PDF", script:"""set -ex
-      conda activate ${ENV_NAME}
-      d2lbook build pdf
-      """
-
-      sh label:"Build Pytorch PDF", script:"""set -ex
-      conda activate ${ENV_NAME}
-      d2lbook build pdf --tab pytorch
-      """
-
-      if (env.BRANCH_NAME == 'release') {
-        sh label:"Release", script:"""set -ex
-        conda activate ${ENV_NAME}
-        d2lbook build pkg
-        d2lbook deploy html pdf slides pkg colab sagemaker --s3 s3://${LANG}-v2.d2l.ai
-        """
-
-      } else {
-        sh label:"Publish", script:"""set -ex
-        conda activate ${ENV_NAME}
-        d2lbook deploy html pdf --s3 s3://preview.d2l.ai/${JOB_NAME}/
-        """
-        if (env.BRANCH_NAME.startsWith("PR-")) {
-            pullRequest.comment("Job ${JOB_NAME}/${BUILD_NUMBER} is complete. \nCheck the results at http://preview.d2l.ai/${JOB_NAME}/")
-        }
-      }
-    }
-  }
-}

README.md

@@ -1,7 +1,5 @@
 # 动手学深度学习（Dive into Deep Learning，D2L.ai）
 
-[![Build Status](http://ci.d2l.ai/job/d2l-zh/job/master/badge/icon)](http://ci.d2l.ai/job/d2l-zh/job/master/)
-
 [第二版：zh.D2L.ai](https://zh.d2l.ai)  | [第一版：zh-v1.D2L.ai](https://zh-v1.d2l.ai/) |  安装和使用书中源代码： [第二版](https://zh.d2l.ai/chapter_installation/index.html) [第一版](https://zh-v1.d2l.ai/chapter_prerequisite/install.html)
 
 <h5 align="center"><i>理解深度学习的最佳方法是学以致用。</i></h5>
@@ -30,11 +28,12 @@
 如果本书对你有帮助，请Star (★) 本仓库或引用本书的英文版：
 
 ```
-@article{zhang2021dive,
+@book{zhang2023dive,
     title={Dive into Deep Learning},
     author={Zhang, Aston and Lipton, Zachary C. and Li, Mu and Smola, Alexander J.},
-    journal={arXiv preprint arXiv:2106.11342},
-    year={2021}
+    publisher={Cambridge University Press},
+    note={\url{https://D2L.ai}},
+    year={2023}
 }
 ```
 

chapter_computational-performance/parameterserver.md

@@ -7,7 +7,7 @@
 
 ## 数据并行训练
 
-让我们回顾一下在分布式架构中数据并行的训练方法，因为在实践中它的实现相对简单，因此本节将排除其他内容只对其进行介绍。由于当今的GPU拥有大量的显存，因此在实际场景中（不包括图深度学习）只有数据并行这种并行训练策略值得推荐。图 :numref:`fig_parameterserver`描述了在 :numref:`sec_multi_gpu`中实现的数据并行的变体。其中的关键是梯度的聚合需要在GPU 0上完成，然后再将更新后的参数广播给所有GPU。
+让我们回顾一下在分布式架构中数据并行的训练方法，因为在实践中它的实现相对简单，因此本节将排除其他内容只对其进行介绍。由于当今的GPU拥有大量的显存，因此在实际场景中（不包括图深度学习）只有数据并行这种并行训练策略值得推荐。图 :numref:`fig_parameterserver`描述了在 :numref:`sec_multi_gpu`中实现的数据并行的变体。其中的关键是梯度的聚合需要在单个GPU（GPU 0）上完成，然后再将更新后的参数广播给所有GPU。
 
 ![左图是单GPU训练；右图是多GPU训练的一个变体：（1）计算损失和梯度，（2）所有梯度聚合在一个GPU上，（3）发生参数更新，并将参数重新广播给所有GPU](../img/ps.svg)
 :label:`fig_parameterserver`
@@ -64,7 +64,7 @@
 ![多机多GPU分布式并行训练](../img/ps-multimachine.svg)
 :label:`fig_ps_multimachine`
 
-以上这些操作似乎都相当简单，而且事实上它们可以在一台机器内高效地执行，但是当我们考虑多台机器时，就会发现中央的参数服务器成为了瓶颈。毕竟，每个服务器的带宽是有限的，因此对$m$个工作节点来说，将所有梯度发送到服务器所需的时间是$\mathcal{O}(m)$。我们也可以通过将参数服务器数量增加到$n$来突破这一障碍。此时，每个服务器只需要存储$\mathcal{O}(1/n)$个参数，因此更新和优化的总时间变为$\mathcal{O}(m/n)$。这两个数字的匹配会产生稳定的伸缩性，而不用在乎我们需要处理多少工作节点。在实际应用中，我们使用同一台机器既作为工作节点还作为服务器。设计说明请参考 :numref:`fig_ps_multips`（技术细节请参考 :cite:`Li.Andersen.Park.ea.2014`）。特别是，确保多台机器只在没有不合理延迟的情况下工作是相当困难的。我们在下面忽略了关于阻塞的细节，只简单介绍一下同步和异步（unsynchronized）更新。
+以上这些操作似乎都相当简单，而且事实上它们可以在一台机器内高效地执行，但是当我们考虑多台机器时，就会发现中央的参数服务器成为了瓶颈。毕竟，每个服务器的带宽是有限的，因此对$m$个工作节点来说，将所有梯度发送到服务器所需的时间是$\mathcal{O}(m)$。我们也可以通过将参数服务器数量增加到$n$来突破这一障碍。此时，每个服务器只需要存储$\mathcal{O}(1/n)$个参数，因此更新和优化的总时间变为$\mathcal{O}(m/n)$。这两个数字的匹配会产生稳定的伸缩性，而不用在乎我们需要处理多少工作节点。在实际应用中，我们使用同一台机器既作为工作节点还作为服务器。设计说明请参考 :numref:`fig_ps_multips`（技术细节请参考 :cite:`Li.Andersen.Park.ea.2014`）。特别是，确保多台机器只在没有不合理延迟的情况下工作是相当困难的。
 
 ![上图：单参数服务器是一个瓶颈，因为它的带宽是有限的；下图：多参数服务器使用聚合带宽存储部分参数](../img/ps-multips.svg)
 :label:`fig_ps_multips`

chapter_multilayer-perceptrons/numerical-stability-and-init.md

@@ -244,7 +244,7 @@ $$
 通常，Xavier初始化从均值为零，方差
 $\sigma^2 = \frac{2}{n_\mathrm{in} + n_\mathrm{out}}$
 的高斯分布中采样权重。
-我们也可以利用Xavier的直觉来选择从均匀分布中抽取权重时的方差。
+我们也可以将其改为选择从均匀分布中抽取权重时的方差。
 注意均匀分布$U(-a, a)$的方差为$\frac{a^2}{3}$。
 将$\frac{a^2}{3}$代入到$\sigma^2$的条件中，将得到初始化值域：
 

chapter_natural-language-processing-pretraining/glove.md

@@ -50,7 +50,7 @@ $$\sum_{i\in\mathcal{V}} \sum_{j\in\mathcal{V}} h(x_{ij}) \left(\mathbf{u}_j^\to
 
 我们也可以从另一个角度来理解GloVe模型。使用 :numref:`subsec_skipgram-global`中的相同符号，设$p_{ij} \stackrel{\mathrm{def}}{=} P(w_j \mid w_i)$为生成上下文词$w_j$的条件概率，给定$w_i$作为语料库中的中心词。 :numref:`tab_glove`根据大量语料库的统计数据，列出了给定单词“ice”和“steam”的共现概率及其比值。
 
-:大型语料库中的词-词共现概率及其比值（根据 :cite:`Pennington.Socher.Manning.2014`中的表1改编）
+大型语料库中的词-词共现概率及其比值（根据 :cite:`Pennington.Socher.Manning.2014`中的表1改编）
 
 |$w_k$=|solid|gas|water|fashion|
 |:--|:-|:-|:-|:-|

chapter_optimization/adadelta.md

@@ -144,7 +144,7 @@ d2l.train_ch11(adadelta, init_adadelta_states(feature_dim),
                {'rho': 0.9}, data_iter, feature_dim);
 ```
 
-为了简洁实现，我们只需使用`Trainer`类中的`adadelta`算法。
+为了简洁实现，我们只需使用高级API中的Adadelta算法。
 
 ```{.python .input}
 d2l.train_concise_ch11('adadelta', {'rho': 0.9}, data_iter)
@@ -197,4 +197,4 @@ d2l.train_concise_ch11(trainer, {'rho': 0.9}, data_iter)
 
 :begin_tab:`paddle`
 [Discussions](https://discuss.d2l.ai/t/11854)
-:end_tab:
+:end_tab:

chapter_optimization/gd.md

@@ -235,7 +235,10 @@ def train_2d(trainer, steps=20, f_grad=None):  #@save
         results.append((x1, x2))
     print(f'epoch {i + 1}, x1: {float(x1):f}, x2: {float(x2):f}')
     return results
+```
 
+```{.python .input}
+#@tab mxnet, tensorflow
 def show_trace_2d(f, results):  #@save
     """显示优化过程中2D变量的轨迹"""
     d2l.set_figsize()
@@ -247,6 +250,19 @@ def show_trace_2d(f, results):  #@save
     d2l.plt.ylabel('x2')
 ```
 
+```{.python .input}
+#@tab pytorch
+def show_trace_2d(f, results):  #@save
+    """显示优化过程中2D变量的轨迹"""
+    d2l.set_figsize()
+    d2l.plt.plot(*zip(*results), '-o', color='#ff7f0e')
+    x1, x2 = d2l.meshgrid(d2l.arange(-5.5, 1.0, 0.1),
+                          d2l.arange(-3.0, 1.0, 0.1), indexing='ij')
+    d2l.plt.contour(x1, x2, f(x1, x2), colors='#1f77b4')
+    d2l.plt.xlabel('x1')
+    d2l.plt.ylabel('x2')
+```
+
 ```{.python .input}
 #@tab paddle
 def train_2d(trainer, steps=20, f_grad=None):  #@save

chapter_optimization/sgd.md

@@ -75,8 +75,8 @@ def f_grad(x1, x2):  # 目标函数的梯度
 def sgd(x1, x2, s1, s2, f_grad):
     g1, g2 = f_grad(x1, x2)
     # 模拟有噪声的梯度
-    g1 += d2l.normal(0.0, 1, (1,))
-    g2 += d2l.normal(0.0, 1, (1,))
+    g1 += d2l.normal(0.0, 1, (1,)).item()
+    g2 += d2l.normal(0.0, 1, (1,)).item()
     eta_t = eta * lr()
     return (x1 - eta_t * g1, x2 - eta_t * g2, 0, 0)
 ```

chapter_preliminaries/pandas.md

@@ -79,7 +79,7 @@ print(inputs)
 ```{.python .input}
 from mxnet import np
 
-X, y = np.array(inputs.values), np.array(outputs.values)
+X, y = np.array(inputs.to_numpy(dtype=float)), np.array(outputs.to_numpy(dtype=float))
 X, y
 ```
 
@@ -87,7 +87,8 @@ X, y
 #@tab pytorch
 import torch
 
-X, y = torch.tensor(inputs.values), torch.tensor(outputs.values)
+X = torch.tensor(inputs.to_numpy(dtype=float))
+y = torch.tensor(outputs.to_numpy(dtype=float))
 X, y
 ```
 
@@ -95,7 +96,8 @@ X, y
 #@tab tensorflow
 import tensorflow as tf
 
-X, y = tf.constant(inputs.values), tf.constant(outputs.values)
+X = tf.constant(inputs.to_numpy(dtype=float))
+y = tf.constant(outputs.to_numpy(dtype=float))
 X, y
 ```
 

ci/docker/Dockerfile.d2l-builder

@@ -0,0 +1,39 @@
+FROM ubuntu:latest
+
+RUN adduser --disabled-password --disabled-login ci
+WORKDIR /home/ci
+
+# Add d2l_job script
+ADD d2l_job.sh .
+RUN chmod +x d2l_job.sh; chown ci d2l_job.sh
+
+# Copy git timesync for caching
+ADD git-timesync /home/ci/.local/bin/
+RUN chmod +x /home/ci/.local/bin/git-timesync
+
+RUN apt-get update && apt-get -y install build-essential git wget
+
+# Install pdf dependencies
+RUN DEBIAN_FRONTEND=noninteractive apt-get install -y texlive-full
+RUN apt-get install -y librsvg2-bin xindy pandoc
+
+# Install fonts
+RUN wget https://raw.githubusercontent.com/d2l-ai/utils/master/install_fonts.sh
+# Remove "sudo " from the fonts script
+RUN sed -i 's/sudo //g' install_fonts.sh
+RUN chmod +x install_fonts.sh; ./install_fonts.sh
+
+RUN apt-get install -y python3 python3-pip python-is-python3
+
+# Allow permissions for pip installations and git-timesync
+RUN chown -R ci:ci /home/ci/.local
+
+USER ci
+
+ENV PATH="/home/ci/.local/bin:$PATH"
+
+# Install d2lbook using pip
+# Install all libraries (cpu) to make sure API reference works for PDF builds
+RUN pip3 install d2lbook==1.0.0 torch torchvision tensorflow mxnet
+
+CMD ["/bin/bash"]

ci/docker/Dockerfile.d2l-zh-mxnet

@@ -0,0 +1,31 @@
+# Use MXNet 1.9.1 (April 2023)
+FROM nvcr.io/nvidia/mxnet:23.04-py3
+
+RUN adduser --disabled-password --disabled-login ci
+WORKDIR /home/ci
+
+# Copy d2l_job script
+ADD d2l_job.sh .
+RUN chmod +x d2l_job.sh; chown ci d2l_job.sh
+
+# Copy git timesync for caching
+ADD git-timesync /home/ci/.local/bin/
+RUN chmod +x /home/ci/.local/bin/git-timesync
+
+# Allow permissions for pip installations and git-timesync
+RUN chown -R ci:ci /home/ci/.local
+
+# Allow write permissions for downloading data to opt/mxnet/data
+RUN chown -R ci:ci /opt
+
+USER ci
+
+ENV PATH="/home/ci/.local/bin:$PATH"
+
+# Install d2lbook using pip
+RUN pip3 install d2lbook==1.0.0
+
+# Python script to print framework versions
+ADD print_versions.py .
+
+CMD ["/bin/bash"]

ci/docker/Dockerfile.d2l-zh-paddle

@@ -0,0 +1,31 @@
+# Use Paddle 2.3.2 (Dec 2022)
+FROM nvcr.io/nvidia/paddlepaddle:22.12-py3
+
+RUN adduser --disabled-password --disabled-login ci
+WORKDIR /home/ci
+
+# Copy d2l_job script
+ADD d2l_job.sh .
+RUN chmod +x d2l_job.sh; chown ci d2l_job.sh
+
+# opencv dependencies required by paddle
+RUN apt-get update && apt-get install libgl1 -y
+
+# Copy git timesync for caching
+ADD git-timesync /home/ci/.local/bin/
+RUN chmod +x /home/ci/.local/bin/git-timesync
+
+# Allow permissions for pip installations and git-timesync
+RUN chown -R ci:ci /home/ci/.local
+
+USER ci
+
+ENV PATH="/home/ci/.local/bin:$PATH"
+
+# Install d2lbook using pip + paddlepaddle dependencies
+RUN pip3 install d2lbook==1.0.0 "opencv-python==4.6.0.66"
+
+# Python script to print framework versions
+ADD print_versions.py .
+
+CMD ["/bin/bash"]

ci/docker/Dockerfile.d2l-zh-tf

@@ -0,0 +1,28 @@
+# Use Tensorflow 2.9.1 (Sept 2023)
+FROM nvcr.io/nvidia/tensorflow:22.09-tf2-py3
+
+RUN adduser --disabled-password --disabled-login ci
+WORKDIR /home/ci
+
+# Copy d2l_job script
+ADD d2l_job.sh .
+RUN chmod +x d2l_job.sh; chown ci d2l_job.sh
+
+# Copy git timesync for caching
+ADD git-timesync /home/ci/.local/bin/
+RUN chmod +x /home/ci/.local/bin/git-timesync
+
+# Allow permissions for pip installations and git-timesync
+RUN chown -R ci:ci /home/ci/.local
+
+USER ci
+
+ENV PATH="/home/ci/.local/bin:$PATH"
+
+# Install d2lbook using pip + dependencies for tensorflow d2l
+RUN pip3 install d2lbook==1.0.0 tensorflow-probability==0.17.0
+
+# Python script to print framework versions
+ADD print_versions.py .
+
+CMD ["/bin/bash"]

ci/docker/Dockerfile.d2l-zh-torch

@@ -0,0 +1,28 @@
+# Use PyTorch 1.12 (May 2022)
+FROM nvcr.io/nvidia/pytorch:22.06-py3
+
+RUN adduser --disabled-password --disabled-login ci
+WORKDIR /home/ci
+
+# Copy d2l_job script
+ADD d2l_job.sh .
+RUN chmod +x d2l_job.sh; chown ci d2l_job.sh
+
+# Copy git timesync for caching
+ADD git-timesync /home/ci/.local/bin/
+RUN chmod +x /home/ci/.local/bin/git-timesync
+
+# Allow permissions for pip installations and git-timesync
+RUN chown -R ci:ci /home/ci/.local
+
+USER ci
+
+ENV PATH="/home/ci/.local/bin:$PATH"
+
+# Install d2lbook using pip + dependencies for torch d2l
+RUN pip3 install d2lbook==1.0.0
+
+# Python script to print framework versions
+ADD print_versions.py .
+
+CMD ["/bin/bash"]

ci/docker/d2l_job.sh

@@ -0,0 +1,68 @@
+#!/bin/bash
+
+date
+echo "Args: $@"
+env
+echo "jobId: $AWS_BATCH_JOB_ID"
+echo "jobQueue: $AWS_BATCH_JQ_NAME"
+echo "computeEnvironment: $AWS_BATCH_CE_NAME"
+
+SOURCE_REF=$1
+WORK_DIR=$2
+COMMAND=$3
+############### NOT USED ATM ##################
+SAVED_OUTPUT=$4
+SAVE_PATH=$5
+###############################################
+REMOTE=$6
+SAFE_TO_USE_SCRIPT=$7
+ORIGINAL_REPO=${8:-'d2l-zh'}
+# TODO @anirudhdagar: hardcode ORIGINAL_ORG
+# Avoid ability to change org by restricting
+# job definition arguments defined in d2l-infra
+# This is only changed for testing purposes
+ORIGINAL_ORG=${9:-'d2l-ai'}
+
+
+# Copy the workflow from master branch
+git clone https://github.com/"$ORIGINAL_ORG"/"$ORIGINAL_REPO".git
+
+WORKFLOW_SCRIPTS="$ORIGINAL_REPO"/.github/workflow_scripts
+if [ -d "$WORKFLOW_SCRIPTS" ]; then
+    cp -R "$ORIGINAL_REPO"/.github/workflow_scripts .
+fi
+
+cd "$ORIGINAL_REPO"
+
+if [ ! -z $REMOTE ]; then
+    git remote set-url origin $REMOTE
+fi
+
+git fetch origin $SOURCE_REF:working
+git checkout working
+
+# Reset modification times for all notebooks using git-timesync
+# We use this to make sure d2lbook build eval caching is valid
+# even after cloning the repo for each run
+# Modification times for original repo files are corrected and are now
+# good for comparing with modification times of build files coming
+# from the S3 bucket
+git timesync *.md **/*.md
+
+# If not safe to use script, we overwrite with the script from master branch
+TRUE=true
+if [[ ${SAFE_TO_USE_SCRIPT,,} != ${TRUE,,} ]]; then
+    if [ -d ../workflow_scripts ]; then
+        rm -rf .github/workflow_scripts
+        mv ../workflow_scripts .github/
+    else
+        echo Not safe to use user provided script, and could not find script from master branches
+        exit 1
+    fi
+fi
+
+cd $WORK_DIR
+/bin/bash -o pipefail -c "eval $COMMAND"
+COMMAND_EXIT_CODE=$?
+
+exit $COMMAND_EXIT_CODE

ci/docker/entry.sh

@@ -0,0 +1,3 @@
+#!/bin/bash
+python3 print_versions.py $1
+/bin/bash

ci/docker/git-timesync

@@ -0,0 +1,288 @@
+#!/bin/sh
+
+# This script has been lifted from https://github.com/tst2005/git-timesync/
+# and all credits for this belong to @tst2005
+
+# shellcheck disable=SC2004,SC3043,SC2155,SC2039
+
+#### About shellcheck
+# I disable:
+# - SC2004 (style): $/${} is unnecessary on arithmetic variables.
+# Because: I prefere `$( $x - $y ))` than `$(( x - y ))`.
+
+# I disable:
+# - SC2039: In POSIX sh, 'local' is undefined.
+# - SC3043: In POSIX sh, 'local' is undefined.
+# Because:
+#   local is too usefull, all modern shell support it.
+#   If you really want to run git-timesync on a strict POSIX shell,
+#   then remove all local prefixes : 's/local //g'.
+#
+# I disable :
+# - SC2155 (warning): Declare and assign separately to avoid masking return values.
+# Because:
+#   It not relevant : the return code is not used.
+#   I prefer `local foo="$(bar)"` than `local foo;foo="$(bar)"`.
+####
+
+set -e
+
+####
+# Author: TsT worldmaster.fr <tst2005@gmail.com>
+#
+# Improvements:
+# - dry-run ("-n" flag)
+# - pass files to check as argument
+# - do not time sync modified files
+# - do not time sync untracked files
+# - performance improvment
+# - be able to apply timesync only on files present in the N last commits
+####
+
+####
+# The original version of this script can be found at: https://gist.github.com/jeffery/1115504
+#
+# Helper script to update the Last modified timestamp of files in a Git SCM
+# Projects working Copy
+#
+# When you clone a Git repository, it sets the timestamp of all the files to the
+# time when you cloned the repository.
+#
+# This becomes a problem when you want the cloned repository, which is part of a 
+# Web application have a proper cacheing mechanism so that it can re-cache files
+# (into a webtree) that have been modified since the last cache.
+#
+# @see http://stackoverflow.com/questions/1964470/whats-the-equivalent-of-use-commit-times-for-git
+#
+# Author: Jeffery Fernandez <jeffery@fernandez.net.au>
+####
+
+showUsage() {
+	echo 'Usage: git-timesync [-n] [-q] [-v] [--] [<paths>...]'
+	echo 'Usage: git-timesync [-n] [-q] [-v] -<N>'
+	echo
+	echo '  -h, --help                 '\
+		'Print this help message'
+	echo '  -n, --dry-run, --dryrun    '\
+		'Perform a dry-run to see which files are OK'\
+		'and which ones need to be synchronized'
+	echo '  -q, --quiet                '\
+		'Quiet mode: drop everything that is OK and show only the files'\
+		'which timestamp needs to be synchronized'
+	echo '  -v, --verbose              '\
+		'Verbose mode: show info for each file (opposite of --quiet)'
+	echo '  -1                         '\
+		'Apply timesync on files present in the last commit'
+	echo '  -23                        '\
+		'Apply timesync on files present in the 23 last commits'
+	echo '  -N                         '\
+		'Apply timesync on files present in the N last commits'\
+		'(with 1 <= N <= 9999)'
+}
+
+# Get the last revision hash of a particular file in the git repository
+getFileLastRevision() {
+	git rev-list HEAD -n 1 -- "$1"
+}
+
+#getFileMTimeByRef() {
+#	git show --pretty=format:%at --abbrev-commit "$1" | head -n 1
+#}
+
+getFileMTimeByPath() {
+	# shellcheck disable=SC2155
+	git rev-list --pretty=format:'date %at' --date-order -n 1 HEAD -- "$1" |
+	(
+		local IFS=" ";
+		# shellcheck disable=SC2034
+		while read -r key value _misc; do
+			[ "$key" != "date" ] || echo "$value";
+		done
+	)
+}
+
+# Extract the actual last modified timestamp of the file and Update the timestamp
+updateFileTimeStamp() {
+	# shellcheck disable=SC2155
+
+	# if target does not exists and it's is not a [dead]link, raise an error
+	if [ ! -e "$1" ] && [ ! -h "$1" ]; then
+		if [ -n "$(git ls-files -t -d -- "$1")" ]; then
+			if $verbose; then echo "?  $1 (deleted)"; fi
+			return
+		fi
+		echo >&2 "ERROR: Unknown bug ?! No such target $1"
+		return 1
+	fi
+
+	local tracked="$(git ls-files -t -c -- "$1")"
+	if [ -z "$tracked" ]; then
+		if $verbose; then echo "?  $1"; fi
+		return
+	fi
+
+	# Extract the last modified timestamp
+	# Get the File last modified time
+	local FILE_MODIFIED_TIME="$(getFileMTimeByPath "$1")"
+	if [ -z "$FILE_MODIFIED_TIME" ]; then
+		echo "?! $1 (not found in git)"
+		return
+	fi
+
+	# Check if the file is modified
+	local uncommited="$(git ls-files -t -dm -- "$1")"
+
+	# for displaying the date in readable format
+	#local FORMATTED_TIMESTAMP="$(date --date="${FILE_MODIFIED_TIME}" +'%d-%m-%Y %H:%M:%S %z')"
+	#local FORMATTED_TIMESTAMP="@${FILE_MODIFIED_TIME}"
+
+	# Modify the last modified timestamp
+	#echo "[$(date -d "$FORMATTED_TIMESTAMP")]: $1"
+	#echo "$FILE_MODIFIED_TIME $1"
+	local current_mtime="$(getmtime "$1")"
+	if $debug; then
+		echo >&2 "DEBUG: $1 (git_time=$FILE_MODIFIED_TIME current_time=$current_mtime delta=$(( ${current_mtime:-0} - ${FILE_MODIFIED_TIME:-0} )))"
+	fi
+	if [ "$current_mtime" = "$FILE_MODIFIED_TIME" ]; then
+		if ${verbose:-true}; then echo "ok $1"; fi
+		return
+	fi
+	if [ -n "$uncommited" ]; then
+		echo "C  $1 (modified, not commited, $(( $current_mtime - $FILE_MODIFIED_TIME ))s recent)"
+		return
+	fi
+	if ${dryrun:-true}; then
+		echo "!! $1 (desync: $(( $current_mtime - $FILE_MODIFIED_TIME ))s, no change)"
+		return
+	fi
+	echo "!! $1 (desync: $(( $current_mtime - $FILE_MODIFIED_TIME ))s, syncing...)"
+	#[ -h "$1" ] && touch -c -h -d "$FORMATTED_TIMESTAMP" -- "$1" || \
+	#touch -c -d "$FORMATTED_TIMESTAMP" -- "$1"
+	unixtime_touch -c -h -- "$1"
+}
+
+
+
+# Make sure we are not running this on a bare Repository
+is_not_base_repo() {
+	case "$(git config core.bare)" in
+		false)	;;
+		true)
+			echo "$(pwd): Cannot run this script on a bare Repository"
+			return 1
+		;;
+		*)	echo "$(pwd): Error appended during core.bare detection. Are you really inside a repository ?"
+			return 1
+	esac
+	return 0
+}
+
+updateFileTimeStampInCwd() {
+	is_not_base_repo || return
+
+	git ls-files -z \
+	| tr '\0' '\n' \
+	| (
+	while read -r file; do
+		if [ -z "$(git ls-files -t -d -- "$file")" ]; then
+			updateFileTimeStamp "${file}"
+		fi
+	done
+	)
+}
+
+timesyncThisFile() {
+	if [ -d "$1" ] && [ ! -h "$1" ]; then # is a real directory (not a symlink to a directory)
+		echo "now inside $1"
+		# shellcheck disable=SC2015
+		( cd -- "$1" && updateFileTimeStampInCwd || true; )
+	else
+		if $need_check_bare; then
+			is_not_base_repo || return 1
+			need_check_bare=false
+		fi
+		updateFileTimeStamp "$1"
+	fi
+}
+
+# ... for Linux ... and MINGW64 (used by Windows GIT Bash)
+linux_unixtime_touch() {
+	# shellcheck disable=SC2155
+	local FORMATTED_TIMESTAMP="@${FILE_MODIFIED_TIME}"
+	touch -d "$FORMATTED_TIMESTAMP" "$@"
+}
+linux_getmtime() {
+	stat -c %Y -- "$1"
+}
+
+# ... for FreeBSD and Mac OS X
+bsd_unixtime_touch() {
+	# shellcheck disable=SC2155
+	local FORMATTED_TIMESTAMP="$(date -j -r "${FILE_MODIFIED_TIME}" +'%Y%m%d%H%M.%S')"
+	touch -t "$FORMATTED_TIMESTAMP" "$@"
+}
+bsd_getmtime() {
+	stat -f %m -- "$1"
+}
+
+################################################################################
+############################## MAIN SCRIPT LOGIC ###############################
+################################################################################
+
+dryrun=false
+verbose=true
+debug=false
+fromrecent=''
+while [ $# -gt 0 ]; do
+	case "$1" in
+		--) shift; break ;;
+		-h|--help) showUsage; exit 0;;
+		-n|--dryrun|--dry-run) dryrun=true ;;
+		-v) verbose=true ;;
+		-q) verbose=false ;;
+		-[1-9]|-[1-9][0-9]|-[1-9][0-9][0-9]|-[1-9][0-9][0-9][0-9]) fromrecent="$1" ;;
+		--debug) debug=true ;;
+		-*) echo >&2 "$0: invalid option $1"; exit 1;;
+		*) break
+	esac
+	shift
+done
+
+# Obtain the Operating System
+case "${GIT_TIMESYNC_FORCE_UNAME:-$(uname)}" in
+	('Linux'|'MINGW64'*)
+		unixtime_touch() { linux_unixtime_touch "$@"; }
+		getmtime() { linux_getmtime "$@"; }
+	;;
+	('Darwin'|'FreeBSD')
+		unixtime_touch() { bsd_unixtime_touch "$@"; }
+		getmtime() { bsd_getmtime "$@"; }
+	;;
+	(*)
+		echo >&2 "Unknown Operating System to perform timestamp update"
+		exit 1
+	;;
+esac
+
+if [ $# -eq 0 ] && [ -z "$fromrecent" ]; then
+	# Loop through and fix timestamps on all files in our checked-out repository
+	updateFileTimeStampInCwd
+else
+	need_check_bare=true
+
+	# Loop through and fix timestamps on all specified files
+	if [ -n "$fromrecent" ]; then
+		git log --format='format:' --name-only "$fromrecent" |
+		sort -u |
+		while read -r file; do
+			[ -n "$file" ] || continue
+			[ -e "$file" ] || continue
+			timesyncThisFile "$file"
+		done
+	else
+		for file in "$@"; do
+			timesyncThisFile "$file"
+		done
+	fi
+fi
+

ci/docker/login_ecr.sh

@@ -0,0 +1,3 @@
+#!/bin/bash
+
+aws ecr get-login-password --region us-west-2 | docker login --username AWS --password-stdin 650140442593.dkr.ecr.us-west-2.amazonaws.com

ci/docker/print_versions.py

@@ -0,0 +1,51 @@
+import os
+import sys
+
+if len(sys.argv) > 1:
+	framework_name = sys.argv[1]
+else:
+	# Assume using d2l-builder docker container
+	# Here all the frameworks are installed and no CUDA support
+	framework_name = None
+
+print("*"*10, "D2L Framework Version Details", "*"*10)
+
+if framework_name:
+	# Print CUDA version
+	print("nvcc --version")
+	print(os.system("nvcc --version"))
+
+if framework_name=="pytorch":
+	# Print PyTorch versions
+	print(f"Framework Name: {framework_name}")
+	import torch; print(f"torch version: {torch.__version__}")
+	import torchvision; print(f"torchvision version: {torchvision.__version__}")
+	import gym; print(f"gym version: {gym.__version__}")
+	import gpytorch; print(f"gpytorch version: {gpytorch.__version__}")
+	import syne_tune; print(f"syne_tune version: {syne_tune.__version__}")
+
+
+if framework_name=="tensorflow":
+	# Print TensorFlow versions
+	print(f"Framework Name: {framework_name}")
+	import tensorflow; print(f"tensorflow version: {tensorflow.__version__}")
+	import tensorflow_probability; print(f"tensorflow_probability version: {tensorflow_probability.__version__}")
+
+if framework_name=="jax":
+	# Print JAX versions
+	print(f"Framework Name: {framework_name}")
+	import jax; print(f"jax version: {jax.__version__}")
+	import jaxlib; print(f"jaxlib version: {jaxlib.__version__}")
+	import flax; print(f"flax version: {flax.__version__}")
+	import tensorflow_datasets; print(f"tensorflow_datasets version: {tensorflow_datasets.__version__}")
+
+if framework_name=="mxnet":
+	# Print MXNet versions
+	print(f"Framework Name: {framework_name}")
+	import mxnet; print(f"MXNet version: {mxnet.__version__}")
+
+
+# Print d2lbook version
+import d2lbook; print(f"d2lbook version: {d2lbook.__version__}")
+
+print("*"*10, "D2L Framework Version Details", "*"*10)

ci/submit-job.py

@@ -0,0 +1,202 @@
+import argparse
+import random
+import os
+import re
+import sys
+import time
+from datetime import datetime
+
+import boto3
+from botocore.compat import total_seconds
+from botocore.config import Config
+
+
+job_type_info = {
+    'ci-cpu': {
+        'job_definition': 'd2l-ci-cpu-builder:2',
+        'job_queue': 'D2L-CI-CPU'
+    },
+    'ci-cpu-push': {
+        'job_definition': 'd2l-ci-cpu-builder-push:7',
+        'job_queue': 'D2L-CI-CPU'
+    },
+    'ci-cpu-release': {
+        'job_definition': 'd2l-ci-cpu-builder-release:1',
+        'job_queue': 'D2L-CI-CPU'
+    },
+    'ci-gpu-torch': {
+        'job_definition': 'd2l-ci-zh-gpu-torch:1',
+        'job_queue': 'D2L-CI-GPU'
+    },
+    'ci-gpu-tf': {
+        'job_definition': 'd2l-ci-zh-gpu-tf:1',
+        'job_queue': 'D2L-CI-GPU'
+    },
+    'ci-gpu-mxnet': {
+        'job_definition': 'd2l-ci-zh-gpu-mxnet:1',
+        'job_queue': 'D2L-CI-GPU'
+    },
+    'ci-gpu-paddle': {
+        'job_definition': 'd2l-ci-zh-gpu-paddle:1',
+        'job_queue': 'D2L-CI-GPU'
+    }
+}
+
+# Create push job types for GPUs with same definitions
+for job_type in list(job_type_info.keys()):
+    if job_type.startswith('ci-gpu'):
+        job_type_info[job_type+'-push'] = job_type_info[job_type]
+        job_type_info[job_type+'-release'] = job_type_info[job_type]
+
+parser = argparse.ArgumentParser(formatter_class=argparse.ArgumentDefaultsHelpFormatter)
+
+parser.add_argument('--profile', help='profile name of aws account.', type=str,
+                    default=None)
+parser.add_argument('--region', help='Default region when creating new connections', type=str,
+                    default='us-west-2')
+parser.add_argument('--name', help='name of the job', type=str, default='d2l-ci')
+parser.add_argument('--job-type', help='type of job to submit.', type=str,
+                    choices=job_type_info.keys(), default='ci-cpu')
+parser.add_argument('--source-ref',
+                    help='ref in d2l-zh main github. e.g. master, refs/pull/500/head',
+                    type=str, default='master')
+parser.add_argument('--work-dir',
+                    help='working directory inside the repo. e.g. scripts/preprocess',
+                    type=str, default='.')
+parser.add_argument('--saved-output',
+                    help='output to be saved, relative to working directory. '
+                         'it can be either a single file or a directory',
+                    type=str, default='None')
+parser.add_argument('--save-path',
+                    help='s3 path where files are saved.',
+                    type=str, default='batch/temp/{}'.format(datetime.now().isoformat()))
+parser.add_argument('--command', help='command to run', type=str,
+                    default='git rev-parse HEAD | tee stdout.log')
+parser.add_argument('--remote',
+                    help='git repo address. https://github.com/d2l-ai/d2l-zh',
+                    type=str, default="https://github.com/d2l-ai/d2l-zh")
+parser.add_argument('--safe-to-use-script',
+                    help='whether the script changes from the actor is safe. We assume it is safe if the actor has write permission to our repo',
+                    action='store_true')
+parser.add_argument('--original-repo', help='name of the repo', type=str, default='d2l-zh')
+parser.add_argument('--wait', help='block wait until the job completes. '
+                    'Non-zero exit code if job fails.', action='store_true')
+parser.add_argument('--timeout', help='job timeout in seconds', default=7200, type=int)
+
+
+args = parser.parse_args()
+
+session = boto3.Session(profile_name=args.profile, region_name=args.region)
+config = Config(
+    retries = dict(
+        max_attempts = 20
+    )
+)
+batch, cloudwatch = [session.client(service_name=sn, config=config) for sn in ['batch', 'logs']]
+
+
+def printLogs(logGroupName, logStreamName, startTime):
+    kwargs = {'logGroupName': logGroupName,
+              'logStreamName': logStreamName,
+              'startTime': startTime,
+              'startFromHead': True}
+
+    lastTimestamp = startTime - 1
+    while True:
+        logEvents = cloudwatch.get_log_events(**kwargs)
+
+        for event in logEvents['events']:
+            lastTimestamp = event['timestamp']
+            timestamp = datetime.utcfromtimestamp(lastTimestamp / 1000.0).isoformat()
+            print('[{}] {}'.format((timestamp + '.000')[:23] + 'Z', event['message']))
+
+        nextToken = logEvents['nextForwardToken']
+        if nextToken and kwargs.get('nextToken') != nextToken:
+            kwargs['nextToken'] = nextToken
+        else:
+            break
+    return lastTimestamp
+
+
+def nowInMillis():
+    endTime = int(total_seconds(datetime.utcnow() - datetime(1970, 1, 1))) * 1000
+    return endTime
+
+
+def main():
+    spin = ['-', '/', '|', '\\', '-', '/', '|', '\\']
+    logGroupName = '/aws/batch/job'
+
+    jobName = re.sub('[^A-Za-z0-9_\-]', '', args.name)[:128]  # Enforce AWS Batch jobName rules
+    jobType = args.job_type
+    jobQueue = job_type_info[jobType]['job_queue']
+    jobDefinition = job_type_info[jobType]['job_definition']
+    wait = args.wait
+
+    safe_to_use_script = 'False'
+    if args.safe_to_use_script:
+        safe_to_use_script = 'True'
+
+    parameters = {
+        'SOURCE_REF': args.source_ref,
+        'WORK_DIR': args.work_dir,
+        'SAVED_OUTPUT': args.saved_output,
+        'SAVE_PATH': args.save_path,
+        'COMMAND': f"\"{args.command}\"",  # wrap command with double quotation mark, so that batch can treat it as a single command
+        'REMOTE': args.remote,
+        'SAFE_TO_USE_SCRIPT': safe_to_use_script,
+        'ORIGINAL_REPO': args.original_repo
+    }
+    kwargs = dict(
+        jobName=jobName,
+        jobQueue=jobQueue,
+        jobDefinition=jobDefinition,
+        parameters=parameters,
+    )
+    if args.timeout is not None:
+        kwargs['timeout'] = {'attemptDurationSeconds': args.timeout}
+    submitJobResponse = batch.submit_job(**kwargs)
+
+    jobId = submitJobResponse['jobId']
+
+    # Export Batch_JobID to Github Actions Environment Variable
+    with open(os.environ['GITHUB_ENV'], 'a') as f:
+        f.write(f'Batch_JobID={jobId}\n')
+    os.environ['batch_jobid'] = jobId
+
+    print('Submitted job [{} - {}] to the job queue [{}]'.format(jobName, jobId, jobQueue))
+
+    spinner = 0
+    running = False
+    status_set = set()
+    startTime = 0
+    logStreamName = None
+    while wait:
+        time.sleep(random.randint(5, 10))
+        describeJobsResponse = batch.describe_jobs(jobs=[jobId])
+        status = describeJobsResponse['jobs'][0]['status']
+        if status == 'SUCCEEDED' or status == 'FAILED':
+            if logStreamName:
+                startTime = printLogs(logGroupName, logStreamName, startTime) + 1
+            print('=' * 80)
+            print('Job [{} - {}] {}'.format(jobName, jobId, status))
+            sys.exit(status == 'FAILED')
+
+        elif status == 'RUNNING':
+            logStreamName = describeJobsResponse['jobs'][0]['container']['logStreamName']
+            if not running:
+                running = True
+                print('\rJob [{}, {}] is RUNNING.'.format(jobName, jobId))
+                if logStreamName:
+                    print('Output [{}]:\n {}'.format(logStreamName, '=' * 80))
+            if logStreamName:
+                startTime = printLogs(logGroupName, logStreamName, startTime) + 1
+        elif status not in status_set:
+            status_set.add(status)
+            print('\rJob [%s - %s] is %-9s... %s' % (jobName, jobId, status, spin[spinner % len(spin)]),)
+            sys.stdout.flush()
+            spinner += 1
+
+
+if __name__ == '__main__':
+    main()

contrib/chapter_recommender-systems/movielens.md

@@ -28,8 +28,8 @@ def read_data_ml100k():
     """读取MovieLens-100k数据集"""
     data_dir = d2l.download_extract('ml-100k')
     names = ['user_id', 'item_id', 'rating', 'timestamp']
-    data = pd.read_csv(os.path.join(data_dir, 'u.data'), '\t', names=names,
-                       engine='python')
+    data = pd.read_csv(os.path.join(data_dir, 'u.data'), sep='\t',
+                       names=names, engine='python')
     num_users = data.user_id.unique().shape[0]
     num_items = data.item_id.unique().shape[0]
     return data, num_users, num_items

static/frontpage/frontpage.html

@@ -203,7 +203,7 @@ a {
     <p class="version"> 跳转<a href="https://zh-v1.d2l.ai/">第一版</a></p>
     <p>面向中文读者的能运行、可讨论的深度学习教科书</p>
     <p>含 PyTorch、NumPy/MXNet、TensorFlow 和 PaddlePaddle 实现</p>
-    <p>被全球 60 多个国家 400 多所大学用于教学</p>
+    <p>被全球 70 多个国家 500 多所大学用于教学</p>
     <p><a class="github-button" href="https://github.com/d2l-ai/d2l-zh" data-icon="octicon-star" data-size="large" data-show-count="true" aria-label="Star d2l-ai/d2l-zh on GitHub">Star</a></p>
   </div>
 </div>
@@ -213,13 +213,18 @@ a {
     <h3>公告</h3>
     <div align="left">
     <ul>
-      <li><b>【重磅升级】</b>
-      第二版纸质书——《动手学深度学习（PyTorch版）》预计于2023年元旦面世。第二版在线内容新增PaddlePaddle实现。
+      <li><b>【重磅升级，<a href="https://raw.githubusercontent.com/d2l-ai/d2l-zh/master/static/frontpage/_images/sales/jd-20230208-zh-6.png">新书榜第一</a>】</b>
+      第二版纸质书——《动手学深度学习（PyTorch版）》（黑白平装版）
+      已在
+      <a href="https://item.jd.com/13628339.html">京东</a>、
+      <a href="https://product.dangdang.com/29511471.html">当当</a>
+      上架。
+      纸质书在内容上与在线版大致相同，但力求在样式、术语标注、语言表述、用词规范、标点以及图、表、章节的索引上符合出版标准和学术规范。
+      第二版在线内容新增PaddlePaddle实现。
       关注本书的<a href="https://github.com/d2l-ai/d2l-zh">中文开源项目</a>和<a href="https://github.com/d2l-ai/d2l-en">英文开源项目</a>以及时获取最新信息。
     </li>
-    <li><b>【购买第一版纸质书（上架4周重印2次，累计3万+册）】</b>
-      纸质书在内容上与在线版大致相同，但力求在样式、术语标注、语言表述、用词规范、标点以及图、表、章节的索引上符合出版标准和学术规范。
-      可以在
+    <li><b>【第一版纸质书】</b>
+      可在
       <a href="https://item.jd.com/12613094.html">京东</a>、
       <a href="http://product.dangdang.com/27872783.html">当当</a>、
       <a href="https://detail.tmall.com/item.htm?id=594937167055">天猫</a>
@@ -232,8 +237,7 @@ a {
       [<a href="https://raw.githubusercontent.com/d2l-ai/d2l-zh/v1/img/frontpage/jd-190715-zh.png">新书榜</a>]
       [<a href="https://zhuanlan.zhihu.com/p/66689123">关于样书</a>]
     </li>
-    <li><b>【免费资源（新增中文版课件）】</b>
-      在校学生和老师可以申请用于本书学习或教学的<a href="http://zh.d2l.ai/aws4learn.html">免费计算资源</a>。
+    <li><b>【免费资源】</b>
       课件、作业、教学视频等资源可参考伯克利“深度学习导论”
       <a href="https://courses.d2l.ai/berkeley-stat-157/syllabus.html">课程大纲</a>
       中的链接（<a href="https://github.com/d2l-ai/berkeley-stat-157/tree/master/slides-zh">中文版课件</a>）。
@@ -257,28 +261,28 @@ a {
     <div class="author-item">
       <img src="./_images/aston.jpg"/>
       <h3><a href="https://astonzhang.github.io/">阿斯顿·张 </a></h3>
-      <p>亚马逊资深科学家</p>
+      <p>亚马逊</p>
     </div>
   </div>
     <div class = "mdl-cell mdl-cell--3-col mdl-cell--top">
     <div class="author-item">
       <img src="./_images/zack.jpg"/>
       <h3><a href="http://zacklipton.com/">扎卡里 C. 立顿</a></h3>
-      <p> 亚马逊科学家，美国卡内基梅隆大学助理教授</p>
+      <p> 美国卡内基梅隆大学、亚马逊</p>
     </div>
   </div>
   <div class = "mdl-cell mdl-cell--3-col mdl-cell--top">
     <div class="author-item">
       <img src="./_images/mu.jpg"/>
       <h3><a href="http://www.cs.cmu.edu/~muli/">李沐</a></h3>
-      <p>亚马逊资深首席科学家</p>
+      <p>亚马逊</p>
     </div>
   </div>
   <div class = "mdl-cell mdl-cell--3-col mdl-cell--top">
     <div class="author-item">
       <img src="./_images/alex.jpg"/>
       <h3><a href="https://alex.smola.org/">亚历山大 J. 斯莫拉</a></h3>
-      <p>亚马逊副总裁/杰出科学家</p>
+      <p>亚马逊</p>
     </div>
   </div>
 
@@ -289,21 +293,21 @@ a {
     <div class="author-item">
       <img src="./_images/brent.jpg"/>
       <h3><a href="https://www.linkedin.com/in/brent-werness-1506471b7/">布伦特 沃尼斯</a></h3>
-      <p>亚马逊资深科学家<br><i><a href="http://d2l.ai/chapter_appendix-mathematics-for-deep-learning/index.html">深度学习的数学</a></i></p>
+      <p>亚马逊<br><i><a href="http://d2l.ai/chapter_appendix-mathematics-for-deep-learning/index.html">深度学习的数学</a></i></p>
     </div>
   </div>
   <div class = "mdl-cell mdl-cell--3-col mdl-cell--top">
     <div class="author-item">
       <img src="./_images/rachel.jpeg"/>
       <h3><a href="https://www.linkedin.com/in/rachelsonghu/">瑞潮儿·胡</a></h3>
-      <p> 亚马逊科学家<br><i><a href="http://d2l.ai/chapter_appendix-mathematics-for-deep-learning/index.html">深度学习的数学</a></i></p>
+      <p> 亚马逊<br><i><a href="http://d2l.ai/chapter_appendix-mathematics-for-deep-learning/index.html">深度学习的数学</a></i></p>
     </div>
   </div>
   <div class = "mdl-cell mdl-cell--3-col mdl-cell--top">
     <div class="author-item">
       <img src="./_images/shuai.jpg"/>
       <h3><a href="https://shuaizhang.tech/">张帅</a></h3>
-      <p>苏黎世联邦理工学院博士后
+      <p>亚马逊
         <br><i><a href="http://d2l.ai/chapter_recommender-systems/index.html">推荐系统</a></i></p>
     </div>
   </div>
@@ -312,7 +316,7 @@ a {
     <div class="author-item">
       <img src="./_images/yi.jpg"/>
       <h3><a href="https://vanzytay.github.io/">郑毅</a></h3>
-      <p>谷歌科学家
+      <p>谷歌
         <br><i><a href="http://d2l.ai/chapter_recommender-systems/index.html">推荐系统</a></i></p>
     </div>
   </div>
@@ -324,42 +328,42 @@ a {
     <div class="author-item">
       <img src="./_images/anirudh.jpg"/>
       <h3><a href="https://github.com/AnirudhDagar">阿尼如 达格</a></h3>
-      <p>印度理工学院罗克分校<br><i>PyTorch改编</i></p>
+      <p>亚马逊<br><i>PyTorch改编</i></p>
     </div>
   </div>
   <div class = "mdl-cell mdl-cell--3-col mdl-cell--top">
     <div class="author-item">
       <img src="./_images/yuan.jpg"/>
       <h3><a href="https://terrytangyuan.github.io/about/">唐源</a></h3>
-      <p>蚂蚁集团高级工程师<br><i>TensorFlow改编</i></p>
+      <p>Akuity<br><i>TensorFlow改编</i></p>
     </div>
   </div>
   <div class = "mdl-cell mdl-cell--3-col mdl-cell--top">
     <div class="author-item">
       <img src="./_images/wugaosheng.jpg"/>
       <h3><a href="https://github.com/w5688414">吴高升</a></h3>
-      <p>百度飞桨高级工程师<br><i>飞桨改编</i></p>
+      <p>百度<br><i>飞桨改编</i></p>
     </div>
   </div>
   <div class = "mdl-cell mdl-cell--3-col mdl-cell--top">
     <div class="author-item">
       <img src="./_images/huliujun.jpg"/>
       <h3><a href="https://github.com/tensorfly-gpu">胡刘俊</a></h3>
-      <p>飞桨开发者技术专家<br><i>飞桨改编</i></p>
+      <p>百度<br><i>飞桨改编</i></p>
     </div>
   </div>
   <div class = "mdl-cell mdl-cell--3-col mdl-cell--top">
     <div class="author-item">
       <img src="./_images/zhangge.jpg"/>
       <h3><a href="https://github.com/Shelly111111">张戈</a></h3>
-      <p>飞桨开发者技术专家<br><i>飞桨改编</i></p>
+      <p>百度<br><i>飞桨改编</i></p>
     </div>
   </div>
   <div class = "mdl-cell mdl-cell--3-col mdl-cell--top">
     <div class="author-item">
       <img src="./_images/xiejiehang.jpg"/>
       <h3><a href="https://github.com/JiehangXie">谢杰航</a></h3>
-      <p>飞桨开发者技术专家<br><i>飞桨改编</i></p>
+      <p>百度<br><i>飞桨改编</i></p>
     </div>
   </div>
 
@@ -370,14 +374,14 @@ a {
     <div class="author-item">
       <img src="./_images/xiaoting.jpg"/>
       <h3><a href="https://github.com/xiaotinghe">何孝霆</a></h3>
-      <p>亚马逊解决方案架构师</p>
+      <p>亚马逊</p>
     </div>
   </div>
   <div class = "mdl-cell mdl-cell--3-col mdl-cell--top">
     <div class="author-item">
       <img src="./_images/rachel.jpeg"/>
       <h3><a href="https://www.linkedin.com/in/rachelsonghu/">瑞潮儿·胡</a></h3>
-      <p>亚马逊科学家</p>
+      <p>亚马逊</p>
     </div>
   </div>
 
@@ -918,7 +922,7 @@ a {
     <!--end of the map-->
 
     <button class="collapsible">
-        <center>[+] <i>点击以显示完整名单</i></center>
+        <center>[+] <i>点击以显示不完整名单</i></center>
     </button>
     <div class="expansion">
     <div class="institute mdl-grid">
@@ -1314,9 +1318,6 @@ a {
     </div> <!-- mdl-grid -->
     </div>
 
-    <h4>如果你使用（或将要使用）本书教学并希望收到免费纸质版图书，
-        请<a href="https://raw.githubusercontent.com/d2l-ai/d2l-en/master/static/frontpage/attachments/hardcopy.txt" target="_blank">联系我们</a>。</h4>
-
     <!-- reverse the changes by google map... -->
     <style>
       a { font-weight: 400; }
@@ -1332,11 +1333,12 @@ a {
       <div class="highlight-python notranslate">
         <div class="highlight">
           <tt>
-            <pre>@article{zhang2021dive,
+            <pre>@book{zhang2023dive,
     title={Dive into Deep Learning},
     author={Zhang, Aston and Lipton, Zachary C. and Li, Mu and Smola, Alexander J.},
-    journal={arXiv preprint arXiv:2106.11342},
-    year={2021}
+    publisher={Cambridge University Press},
+    note={\url{https://D2L.ai}},
+    year={2023}
 }</pre>
           </tt>
         </div>