Add support to ONNX models as new engine provider for Deep Learning models

.github/workflows/ci-build-manual-onnx.yml

@@ -0,0 +1,79 @@
+name: Build and push an ONNX docker image
+
+# This workflow builds the lightweight ONNX/Wapiti-only Docker image
+# (no Python/DeLFT/TensorFlow dependencies)
+
+on:
+  push:
+    branches:
+      - feature/onnx-models
+  workflow_dispatch:
+    inputs:
+      custom_tag:
+        type: string
+        description: Docker image tag
+        required: true
+        default: "latest-onnx"
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+
+    steps:
+    - uses: actions/checkout@v5
+      with:
+        fetch-tags: true
+        fetch-depth: 0
+    - name: Set up JDK 21
+      uses: actions/setup-java@v5
+      with:
+        java-version: '21'
+        distribution: 'temurin'
+        cache: 'gradle'
+    - name: Build with Gradle
+      run: ./gradlew build -x test
+
+  docker-build-onnx:
+    needs: [ build ]
+    runs-on: ubuntu-latest
+
+    steps:
+      - name: Create more disk space
+        run: sudo rm -rf /usr/share/dotnet && sudo rm -rf /opt/ghc && sudo rm -rf "/usr/local/share/boost" && sudo rm -rf "$AGENT_TOOLSDIRECTORY"
+      - uses: actions/checkout@v5
+      - name: Build and push
+        id: docker_build
+        uses: mr-smithers-excellent/docker-build-push@v5
+        with:
+          username: ${{ secrets.DOCKERHUB_USERNAME_LFOPPIANO }}
+          password: ${{ secrets.DOCKERHUB_TOKEN_LFOPPIANO }}
+          image: lfoppiano/grobid
+          registry: docker.io
+          pushImage: true
+          tags: latest-onnx, ${{ github.event.inputs.custom_tag || github.sha }}
+          dockerfile: Dockerfile.onnx
+      - name: Image digest
+        run: echo ${{ steps.docker_build.outputs.digest }}
+      - name: Docker Image Summary
+        run: |
+          echo "## 🐳 Docker Image Uploaded Successfully" >> $GITHUB_STEP_SUMMARY
+          echo "" >> $GITHUB_STEP_SUMMARY
+          echo "**Image Details:**" >> $GITHUB_STEP_SUMMARY
+          echo "- **Registry:** docker.io" >> $GITHUB_STEP_SUMMARY
+          echo "- **Image:** lfoppiano/grobid" >> $GITHUB_STEP_SUMMARY
+          echo "- **Type:** ONNX/Wapiti only (lightweight, no Python/DeLFT)" >> $GITHUB_STEP_SUMMARY
+          echo "- **Tags:**" >> $GITHUB_STEP_SUMMARY
+          echo "  - \`latest-onnx\`" >> $GITHUB_STEP_SUMMARY
+          echo "  - \`${{ github.event.inputs.custom_tag || github.sha }}\`" >> $GITHUB_STEP_SUMMARY
+          echo "- **Digest:** \`${{ steps.docker_build.outputs.digest }}\`" >> $GITHUB_STEP_SUMMARY
+          echo "" >> $GITHUB_STEP_SUMMARY
+          echo "**Features:**" >> $GITHUB_STEP_SUMMARY
+          echo "- ONNX Runtime for deep learning models (CPU only)" >> $GITHUB_STEP_SUMMARY
+          echo "- Wapiti CRF for traditional models" >> $GITHUB_STEP_SUMMARY
+          echo "- No Python, TensorFlow, or DeLFT dependencies" >> $GITHUB_STEP_SUMMARY
+          echo "" >> $GITHUB_STEP_SUMMARY
+          echo "**Usage:**" >> $GITHUB_STEP_SUMMARY
+          echo "\`\`\`bash" >> $GITHUB_STEP_SUMMARY
+          echo "docker pull lfoppiano/grobid:latest-onnx" >> $GITHUB_STEP_SUMMARY
+          echo "docker run -t --rm --init -p 8070:8070 -p 8071:8071 lfoppiano/grobid:latest-onnx" >> $GITHUB_STEP_SUMMARY
+          echo "\`\`\`" >> $GITHUB_STEP_SUMMARY

.gitignore

@@ -89,3 +89,5 @@ Dockerfile.dataseer
 Dockerfile.software
 Dockerfile.datastet
 .run
+
+.kotlin

Dockerfile.crf

@@ -49,8 +49,9 @@ RUN rm -rf grobid-home/lib/lin-32
 RUN rm -rf grobid-home/lib/win-*
 RUN rm -rf grobid-home/lib/mac-64
 
-# cleaning Delft models
-RUN rm -rf grobid-home/models/*-BidLSTM_CRF*
+# cleaning Delft and ONNX models
+RUN rm -rf grobid-home/models/*-BidLSTM*
+RUN rm -rf grobid-home/models/*.onnx
 
 ENV GROBID_SERVICE_OPTS="-Djava.library.path=grobid-home/lib/lin-64:grobid-home/lib/lin-64/jep"
 

Dockerfile.onnx

@@ -0,0 +1,117 @@
+## Docker GROBID image using ONNX models and/or Wapiti CRF models
+## This is a lightweight image without Python/TensorFlow/DeLFT/JEP dependencies
+## Uses ONNX Runtime (CPU only)
+
+## See https://grobid.readthedocs.io/en/latest/Grobid-docker/
+
+## usage example with version 0.8.0:
+## docker build -t grobid/grobid:0.8.0-onnx --build-arg GROBID_VERSION=0.8.0 --file Dockerfile.onnx .
+
+## run:
+## docker run -t --rm --init -p 8070:8070 -p 8071:8071 grobid/grobid:0.8.0-onnx
+
+# -------------------
+# build builder image
+# -------------------
+
+FROM eclipse-temurin:21-jdk AS builder
+
+USER root
+
+RUN apt-get update && \
+    apt-get -y upgrade && \
+    apt-get -y --no-install-recommends install unzip git python3 python3-pip
+
+WORKDIR /opt/grobid-source
+
+# gradle
+COPY gradle/ ./gradle/
+COPY gradlew ./
+COPY gradle.properties ./
+COPY build.gradle ./
+COPY settings.gradle ./
+
+# git
+COPY .git/ ./.git
+
+# source
+COPY grobid-home/ ./grobid-home/
+COPY grobid-core/ ./grobid-core/
+COPY grobid-service/ ./grobid-service/
+COPY grobid-trainer/ ./grobid-trainer/
+
+# cleaning unused native libraries before packaging
+RUN rm -rf grobid-home/pdf2xml
+RUN rm -rf grobid-home/pdfalto/lin-32
+RUN rm -rf grobid-home/pdfalto/mac-64
+RUN rm -rf grobid-home/pdfalto/mac_arm-64
+RUN rm -rf grobid-home/pdfalto/win-*
+RUN rm -rf grobid-home/lib/lin-32
+RUN rm -rf grobid-home/lib/win-*
+RUN rm -rf grobid-home/lib/mac-64
+RUN rm -rf grobid-home/lib/lin-64/jep
+
+# Use ONNX configuration (no DeLFT models)
+RUN rm grobid-home/config/grobid.yaml && \
+    mv grobid-home/config/grobid-onnx.yaml grobid-home/config/grobid.yaml
+
+RUN ./gradlew clean assemble --no-daemon --info --stacktrace
+
+# Preload embeddings in raw float32 format for ONNX inference
+# Using standalone script that doesn't require DeLFT
+RUN pip3 install --no-cache-dir --break-system-packages lmdb requests
+COPY grobid-home/scripts/preload_embeddings_standalone.py .
+COPY grobid-home/config/resources-registry.json .
+RUN python3 preload_embeddings_standalone.py --registry ./resources-registry.json
+
+WORKDIR /opt/grobid
+RUN unzip -o /opt/grobid-source/grobid-service/build/distributions/grobid-service-*.zip && \
+    mv grobid-service* grobid-service
+RUN unzip -o /opt/grobid-source/grobid-home/build/distributions/grobid-home-*.zip && \
+    chmod -R 755 /opt/grobid/grobid-home/pdfalto
+
+# Move preloaded embeddings to final location
+RUN mkdir -p /opt/grobid/data/db && \
+    mv /opt/grobid-source/data/db/* /opt/grobid/data/db/
+
+# -------------------
+# build runtime image
+# -------------------
+
+FROM eclipse-temurin:21-jre
+
+# setting locale
+ENV LANG=C.UTF-8
+
+# Install minimal runtime dependencies
+RUN apt-get update && \
+    apt-get -y --no-install-recommends install \
+    libxml2 libfontconfig \
+    && apt-get clean \
+    && rm -rf /var/lib/apt/lists/*
+
+WORKDIR /opt/grobid
+
+COPY --from=builder /opt/grobid .
+
+# Add Tini
+ENV TINI_VERSION=v0.19.0
+ADD https://github.com/krallin/tini/releases/download/${TINI_VERSION}/tini /tini
+RUN chmod +x /tini
+ENTRYPOINT ["/tini", "-s", "--"]
+
+WORKDIR /opt/grobid
+
+ENV JAVA_OPTS="-Xmx4g --add-opens=java.base/java.lang=ALL-UNNAMED --add-opens=java.base/java.nio=ALL-UNNAMED --add-opens=java.base/sun.nio.ch=ALL-UNNAMED --add-opens=java.base/java.io=ALL-UNNAMED"
+
+
+CMD ["./grobid-service/bin/grobid-service"]
+
+ARG GROBID_VERSION
+
+LABEL \
+    authors="The contributors" \
+    org.label-schema.name="GROBID" \
+    org.label-schema.description="Image with GROBID service (ONNX/Wapiti only, no DeLFT)" \
+    org.label-schema.url="https://github.com/kermitt2/grobid" \
+    org.label-schema.version=${GROBID_VERSION}