feat: Add llama (#23)

* feat: Add llama Signed-off-by: Ce Gao <[email protected]> * chore: Add git for git clone in requirements Signed-off-by: Ce Gao <[email protected]> * fix: Disable PR trigger Signed-off-by: Ce Gao <[email protected]> --------- Signed-off-by: Ce Gao <[email protected]>
tensorchord · May 24, 2023 · 18f2a54 · 18f2a54
1 parent 126af31
commit 18f2a54
Show file tree

Hide file tree

Showing 5 changed files with 106 additions and 14 deletions.
diff --git a/.github/workflows/docker-publish.yml b/.github/workflows/docker-publish.yml
@@ -10,22 +10,23 @@ on:
     types: [created]
   # Allows you to run this workflow manually from the Actions tab
   workflow_dispatch:
-  pull_request:
-    paths:
-    - '.github/workflows/docker-publish.yml'
 
 concurrency:
   group: ${{ github.ref }}-${{ github.workflow }}
   cancel-in-progress: true
 
 env:
   REGISTRY: docker.io
-  IMAGE_NAME: modelzai/llm
-
 
 jobs:
   build:
-
+    strategy:
+      matrix:
+        image:
+          - name: modelzai/llm-chatglm-6b
+            dockerfile: ./images/base/Dockerfile
+          - name: modelzai/llm-llama-7b
+            dockerfile: ./images/llama/Dockerfile
     runs-on: ubuntu-latest
     permissions:
       contents: read
@@ -50,7 +51,7 @@ jobs:
         id: meta
         uses: docker/metadata-action@v4
         with:
-          images: ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}
+          images: ${{ env.REGISTRY }}/${{ matrix.image.name }}
 
       # Build and push Docker image with Buildx (don't push on PR)
       # https://github.com/docker/build-push-action
@@ -59,6 +60,7 @@ jobs:
         uses: docker/build-push-action@v4
         with:
           context: .
+          file: ${{ matrix.image.dockerfile }}
           push: ${{ github.event_name != 'pull_request' }}
           tags: ${{ steps.meta.outputs.tags }}
           labels: ${{ steps.meta.outputs.labels }}

diff --git a/.github/workflows/gcr.yml b/.github/workflows/gcr.yml
@@ -10,22 +10,23 @@ on:
     types: [created]
   # Allows you to run this workflow manually from the Actions tab
   workflow_dispatch:
-  pull_request:
-    paths:
-    - '.github/workflows/gcr.yml'
 
 concurrency:
   group: ${{ github.ref }}-${{ github.workflow }}
   cancel-in-progress: true
 
 env:
   REGISTRY: us-central1-docker.pkg.dev/nth-guide-378813
-  IMAGE_NAME: modelzai/llm
-
 
 jobs:
   build:
-
+    strategy:
+      matrix:
+        image:
+          - name: modelzai/llm-chatglm-6b
+            dockerfile: ./images/base/Dockerfile
+          - name: modelzai/llm-llama-7b
+            dockerfile: ./images/llama/Dockerfile
     runs-on: ubuntu-latest
     permissions:
       contents: read
@@ -54,7 +55,7 @@ jobs:
         id: meta
         uses: docker/metadata-action@v4
         with:
-          images: ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}
+          images: ${{ env.REGISTRY }}/${{ matrix.image.name }}
 
       # Build and push Docker image with Buildx (don't push on PR)
       # https://github.com/docker/build-push-action
@@ -63,6 +64,7 @@ jobs:
         uses: docker/build-push-action@v4
         with:
           context: .
+          file: ${{ matrix.image.dockerfile }}
           push: ${{ github.event_name != 'pull_request' }}
           tags: ${{ steps.meta.outputs.tags }}
           labels: ${{ steps.meta.outputs.labels }}

diff --git a/Dockerfile → images/base/Dockerfile b/Dockerfile → images/base/Dockerfile
diff --git a/images/llama-7b/Dockerfile b/images/llama-7b/Dockerfile
@@ -0,0 +1,76 @@
+ARG base=nvidia/cuda:11.6.2-cudnn8-runtime-ubuntu20.04
+
+FROM ${base}
+
+ENV DEBIAN_FRONTEND=noninteractive LANG=en_US.UTF-8 LC_ALL=en_US.UTF-8
+ENV PATH /opt/conda/bin:$PATH
+
+ENV MODELZ_MODEL=decapoda-research/llama-7b-hf
+ENV MODELZ_TOKENIZER=decapoda-research/llama-7b-hf
+
+ARG MOSEC_PORT=8080
+ENV MOSEC_PORT=${MOSEC_PORT}
+
+ARG CONDA_VERSION=py310_22.11.1-1
+
+RUN apt update && \
+    apt install -y --no-install-recommends \
+        wget \
+        git \
+        build-essential \
+        ca-certificates && \
+    rm -rf /var/lib/apt/lists/*
+
+RUN set -x && \
+    UNAME_M="$(uname -m)" && \
+    if [ "${UNAME_M}" = "x86_64" ]; then \
+        MINICONDA_URL="https://repo.anaconda.com/miniconda/Miniconda3-${CONDA_VERSION}-Linux-x86_64.sh"; \
+        SHA256SUM="00938c3534750a0e4069499baf8f4e6dc1c2e471c86a59caa0dd03f4a9269db6"; \
+    elif [ "${UNAME_M}" = "s390x" ]; then \
+        MINICONDA_URL="https://repo.anaconda.com/miniconda/Miniconda3-${CONDA_VERSION}-Linux-s390x.sh"; \
+        SHA256SUM="a150511e7fd19d07b770f278fb5dd2df4bc24a8f55f06d6274774f209a36c766"; \
+    elif [ "${UNAME_M}" = "aarch64" ]; then \
+        MINICONDA_URL="https://repo.anaconda.com/miniconda/Miniconda3-${CONDA_VERSION}-Linux-aarch64.sh"; \
+        SHA256SUM="48a96df9ff56f7421b6dd7f9f71d548023847ba918c3826059918c08326c2017"; \
+    elif [ "${UNAME_M}" = "ppc64le" ]; then \
+        MINICONDA_URL="https://repo.anaconda.com/miniconda/Miniconda3-${CONDA_VERSION}-Linux-ppc64le.sh"; \
+        SHA256SUM="4c86c3383bb27b44f7059336c3a46c34922df42824577b93eadecefbf7423836"; \
+    fi && \
+    wget "${MINICONDA_URL}" -O miniconda.sh -q && \
+    echo "${SHA256SUM} miniconda.sh" > shasum && \
+    if [ "${CONDA_VERSION}" != "latest" ]; then sha256sum --check --status shasum; fi && \
+    mkdir -p /opt && \
+    bash miniconda.sh -b -p /opt/conda && \
+    rm miniconda.sh shasum && \
+    ln -s /opt/conda/etc/profile.d/conda.sh /etc/profile.d/conda.sh && \
+    echo ". /opt/conda/etc/profile.d/conda.sh" >> ~/.bashrc && \
+    echo "conda activate base" >> ~/.bashrc && \
+    find /opt/conda/ -follow -type f -name '*.a' -delete && \
+    find /opt/conda/ -follow -type f -name '*.js.map' -delete && \
+    /opt/conda/bin/conda clean -afy
+
+RUN conda create -n envd python=3.9
+
+ENV ENVD_PREFIX=/opt/conda/envs/envd/bin
+
+RUN update-alternatives --install /usr/bin/python python ${ENVD_PREFIX}/python 1 && \
+    update-alternatives --install /usr/bin/python3 python3 ${ENVD_PREFIX}/python3 1 && \
+    update-alternatives --install /usr/bin/pip pip ${ENVD_PREFIX}/pip 1 && \
+    update-alternatives --install /usr/bin/pip3 pip3 ${ENVD_PREFIX}/pip3 1
+
+COPY ./images/llama/requirements.txt /
+
+RUN pip install -r requirements.txt
+
+RUN mkdir -p /workspace
+
+COPY main.py workspace/
+
+WORKDIR /workspace
+
+RUN python main.py --dry-run
+
+# disable huggingface update check (could be very slow)
+ENV HF_HUB_OFFLINE=true
+
+ENTRYPOINT [ "uvicorn", "main:app", "--host", "0.0.0.0", "--port", "8080", "--workers", "1" ]
diff --git a/images/llama-7b/requirements.txt b/images/llama-7b/requirements.txt
@@ -0,0 +1,12 @@
+msgpack
+mosec
+torch --extra-index-url https://download.pytorch.org/whl/cu116
+diffusers[torch]
+# Ref https://github.com/huggingface/transformers/issues/22222
+transformers @ git+https://github.com/zphang/transformers.git@68d640f7c368bcaaaecfc678f11908ebbd3d6176
+sentencepiece
+cpm_kernels
+accelerate
+llmspec
+falcon
+uvicorn