Dockerfile.train_bi

FROM nvidia/cuda:12.1.1-devel-ubuntu22.04

RUN apt-get update && \
    apt-get install -y --no-install-recommends \
    git wget build-essential \
    ca-certificates libjpeg-dev libpng-dev && \
    rm -rf /var/lib/apt/lists/*

ENV CONDA_DIR /opt/conda
ENV PATH $CONDA_DIR/bin:$PATH
ENV CUDA_HOME /usr/local/cuda

RUN wget --quiet https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh -O /tmp/miniconda.sh && \
    /bin/bash /tmp/miniconda.sh -b -p $CONDA_DIR && \
    rm /tmp/miniconda.sh && \
    $CONDA_DIR/bin/conda clean -ya

RUN conda create -n docker python=3.11 -y

WORKDIR /app

COPY requirements_lang.txt /app/requirements_lang.txt

RUN conda run -n docker conda install -c pytorch -c nvidia \
    pytorch torchvision torchaudio pytorch-cuda=12.1 -y && \
    conda run -n docker pip install --no-cache-dir -r /app/requirements_lang.txt && \
    conda run -n docker pip install --no-cache-dir pandas deepspeed sentencepiece tqdm

COPY . /app

RUN conda run -n docker pip install -e .

ENTRYPOINT ["conda", "run", "-n", "docker"]

CMD ["torchrun", "--nproc_per_node", "4", \
     "-m", "FlagEmbedding.baai_general_embedding.finetune.run", \
     "--output_dir", "output_dir/ckpt_bi_encoder", \
     "--model_name_or_path", "BAAI/bge-m3", \
     "--train_data", "data/bi/train_1024_chunks.jsonl", \
     "--learning_rate", "1e-5", \
     "--num_train_epochs", "20", \
     "--per_device_train_batch_size", "2", \
     "--dataloader_drop_last", "True", \
     "--deepspeed", "FlagEmbedding/BGE_M3/ds_config.json", \
     "--normlized", "True", \
     "--temperature", "0.02", \
     "--warmup_ratio", "0.1", \
     "--query_max_len", "128", \
     "--passage_max_len", "1024", \
     "--train_group_size", "4", \
     "--negatives_cross_device", \
     "--logging_steps", "10", \
     "--fp16", \
     "--save_steps", "1000"]