Dockerfile

FROM nvidia/cuda:12.1.1-cudnn8-runtime-ubuntu22.04

ENV DEBIAN_FRONTEND=noninteractive
WORKDIR /home/user/app

ENV OMP_NUM_THREADS=1 \
    TOKENIZERS_PARALLELISM=false \
    PYTHONUNBUFFERED=1 \
    HF_HOME=/home/user/.cache/huggingface \
    TRANSFORMERS_CACHE=/home/user/.cache/huggingface/transformers \
    HF_DATASETS_CACHE=/home/user/.cache/huggingface/datasets \
    OFFLOAD_DIR=/home/user/app/offload \
    MERGED_MODEL_DIR=/home/user/app/merged-model \
    QUANTIZE=none \                  # <── default: no quantization
    USE_ADAPTER_INFERENCE=1 \
    FORCE_REMERGE=0 \
    ALLOW_RUNTIME_MERGE=0 \
    NVIDIA_VISIBLE_DEVICES=all \
    NVIDIA_DRIVER_CAPABILITIES=compute,utility

# System deps
RUN apt-get update && apt-get install -y --no-install-recommends \
    python3 python3-pip git git-lfs ca-certificates && \
    rm -rf /var/lib/apt/lists/* && \
    git lfs install

# Torch stack aligned with transformers 4.44
RUN python3 -m pip install --upgrade pip
RUN python3 -m pip install --no-cache-dir \
    --index-url https://download.pytorch.org/whl/cu121 \
    torch==2.3.1 torchvision==0.18.1 torchaudio==2.3.1

# Only CPU FAISS (not GPU)
COPY requirements.txt .
RUN python3 -m pip install --no-cache-dir -r requirements.txt

COPY . .

RUN mkdir -p "$OFFLOAD_DIR" "$MERGED_MODEL_DIR" "$HF_HOME" "$TRANSFORMERS_CACHE"

EXPOSE 7860
CMD ["python3", "app.py"]