Add pytorch/inference/gpu/2.3.1/transformers/4.46.0/py311 (WIP)

- Include missing `requirements.txt` installation in `entrypoint.sh`
(required to install custom dependencies with custom models)
- Fix Python 3.11 installation as it was not properly installed and
Python 3.10 was used instead
- Use `uv` to install the dependencies as it's way faster than default
`pip`
- Also `uv` is able to successfully install `kenlm` which is a
`transformers` dependency that `pip` is not able to install when
building the `Dockerfile`
- Tested with some of the latest models that those bumped dependencies
support as Gemma2, Llama3.2, StableDiffusion 3.5, and much more

containers/pytorch/inference/gpu/2.3.1/transformers/4.46.0/py311/Dockerfile

@@ -0,0 +1,65 @@
+FROM nvidia/cuda:12.1.0-devel-ubuntu22.04
+SHELL ["/bin/bash", "-c"]
+
+LABEL maintainer="Hugging Face"
+
+ENV DEBIAN_FRONTEND=noninteractive
+
+WORKDIR /app
+
+# Install required dependencies
+RUN apt-get update && \
+    apt-get install software-properties-common -y && \
+    add-apt-repository ppa:deadsnakes/ppa && \
+    apt-get -y upgrade --only-upgrade systemd openssl cryptsetup && \
+    apt-get install -y \
+    build-essential \
+    bzip2 \
+    curl \
+    git \
+    git-lfs \
+    tar \
+    gcc \
+    g++ \
+    cmake \
+    libprotobuf-dev \
+    protobuf-compiler \
+    python3.11 \
+    python3.11-dev \
+    libsndfile1-dev \
+    ffmpeg && \
+    rm -rf /var/lib/apt/lists/*
+
+# Set Python 3.11 as the default python version
+RUN update-alternatives --install /usr/bin/python3 python3 /usr/bin/python3.11 1 && \
+    ln -sf /usr/bin/python3.11 /usr/bin/python
+
+# Install pip from source
+RUN curl -O https://bootstrap.pypa.io/get-pip.py && \
+    python get-pip.py && \
+    rm get-pip.py
+
+# Install uv and upgrade setuptools
+RUN pip install --upgrade uv && \
+    uv pip install --upgrade setuptools --no-cache-dir --system
+
+# Hugging Face Inference Toolkit
+ARG HF_INFERENCE_TOOLKIT_VERSION=bump-dependencies
+ARG HF_INFERENCE_TOOLKIT_URL=git+https://github.com/huggingface/huggingface-inference-toolkit.git@${HF_INFERENCE_TOOLKIT_VERSION}
+RUN uv pip install --upgrade "${HF_INFERENCE_TOOLKIT_URL}#egg=huggingface-inference-toolkit[torch,diffusers,st,google]" --no-cache-dir --system
+
+ENV HF_HUB_ENABLE_HF_TRANSFER="1"
+
+# Install Google CLI single command
+RUN echo "deb [signed-by=/usr/share/keyrings/cloud.google.gpg] https://packages.cloud.google.com/apt cloud-sdk main" \
+    | tee -a /etc/apt/sources.list.d/google-cloud-sdk.list && \
+    curl https://packages.cloud.google.com/apt/doc/apt-key.gpg \
+    | apt-key --keyring /usr/share/keyrings/cloud.google.gpg add - && \
+    apt-get update -y && \
+    apt-get install google-cloud-sdk -y && \
+    apt-get clean autoremove --yes && \
+    rm -rf /var/lib/{apt,dpkg,cache,log}
+
+# Copy entrypoint and change permissions
+COPY --chmod=0755 containers/pytorch/inference/gpu/2.3.1/transformers/4.46.0/py311/entrypoint.sh entrypoint.sh
+ENTRYPOINT ["bash", "-c", "./entrypoint.sh"]

containers/pytorch/inference/gpu/2.3.1/transformers/4.46.0/py311/entrypoint.sh

@@ -0,0 +1,44 @@
+#!/bin/bash
+
+# Define the default port
+PORT=5000
+
+# Check if AIP_MODE is set and adjust the port for Vertex AI
+if [[ ! -z "${AIP_MODE}" ]]; then
+    PORT=${AIP_HTTP_PORT}
+fi
+
+# Check if MODEL_ID starts with "gcs://"
+if [[ $AIP_STORAGE_URI == gs://* ]]; then
+    echo "AIP_STORAGE_URI set and starts with 'gs://', proceeding to download from GCS."
+    echo "AIP_STORAGE_URI: $AIP_STORAGE_URI"
+
+    # Define the target directory
+    TARGET_DIR="/opt/huggingface/model"
+    mkdir -p "$TARGET_DIR"
+
+    # Use gsutil to copy the content from GCS to the target directory
+    echo "Running: gsutil -m cp -e -r "$AIP_STORAGE_URI/*" "$TARGET_DIR""
+    gsutil -m cp -e -r "$AIP_STORAGE_URI/*" "$TARGET_DIR"
+
+    # Check if gsutil command was successful
+    if [ $? -eq 0 ]; then
+        echo "Model downloaded successfully to ${TARGET_DIR}."
+        # Update MODEL_ID to point to the local directory
+        echo "Updating MODEL_ID to point to the local directory."
+        export HF_MODEL_DIR="$TARGET_DIR"
+        export AIP_STORAGE_URI=""
+    else
+        echo "Failed to download model from GCS."
+        exit 1
+    fi
+
+    # Check if requirements.txt exists and if so install dependencies
+    if [ -f "${HF_MODEL_DIR}/requirements.txt" ]; then
+        echo "Installing custom dependencies from ${HF_MODEL_DIR}/requirements.txt"
+        uv pip install -r ${HF_MODEL_DIR}/requirements.txt --no-cache-dir --system
+    fi
+fi
+
+# Start the server
+uvicorn huggingface_inference_toolkit.webservice_starlette:app --host 0.0.0.0 --port ${PORT}