hkz
5ea1bf2450
* Add VILA1.5 in video caption * update VILA1.5 in video_caption * add get_video_path_list back & update the accelerate version & auto load the vila quant_path & download models in the main process * add exception handling & add force_ascii=False * fix the NCCL timeout & use logger * update utils * Update video splitting * Update pre-filtering * Update video caption * Add caption_rewrite.py * Add VideoCLIPXL * Add beautiful prompt demo * fix test * Update Dockerfile.ds * update VideoCLIPXL & add filter_meta_train.py * update stage3 * update README.md * update README_zh-CN.md * update README * update requirements * fix stage_3_video_recaptioning.sh * update Beautiful Prompt * update README * update vila_video_recaptioning.py & fix the empty gather result
68 lines
3.6 KiB
Docker
68 lines
3.6 KiB
Docker
FROM nvidia/cuda:11.8.0-devel-ubuntu22.04
|
|
ENV DEBIAN_FRONTEND noninteractive
|
|
|
|
RUN rm -r /etc/apt/sources.list.d/
|
|
|
|
RUN apt-get update -y && apt-get install -y \
|
|
libgl1 libglib2.0-0 google-perftools \
|
|
sudo wget git git-lfs vim tig pkg-config libcairo2-dev \
|
|
telnet curl net-tools iputils-ping wget jq \
|
|
python3-pip python-is-python3 python3.10-venv tzdata lsof && \
|
|
rm -rf /var/lib/apt/lists/*
|
|
RUN pip3 install --upgrade pip -i https://mirrors.aliyun.com/pypi/simple/
|
|
|
|
# add all extensions
|
|
RUN apt-get update -y && apt-get install -y zip && \
|
|
rm -rf /var/lib/apt/lists/*
|
|
RUN pip install wandb tqdm GitPython==3.1.32 Pillow==9.5.0 setuptools --upgrade -i https://mirrors.aliyun.com/pypi/simple/
|
|
|
|
RUN pip install torch==2.4.0 torchvision==0.19.0 torchaudio==2.4.0 --index-url https://download.pytorch.org/whl/cu118
|
|
RUN pip install xformers==0.0.27.post2 --index-url https://download.pytorch.org/whl/cu118
|
|
|
|
# install VILA (video-caption)
|
|
RUN wget https://pai-aigc-photog.oss-cn-hangzhou.aliyuncs.com/easyanimate/package/vila-1.0.0-torch2.4.0-py3-none-any.whl
|
|
RUN mv vila-1.0.0-torch2.4.0-py3-none-any.whl vila-1.0.0-py3-none-any.whl
|
|
RUN pip install vila-1.0.0-py3-none-any.whl --extra-index-url https://download.pytorch.org/whl/cu118
|
|
RUN wget https://pai-aigc-photog.oss-cn-hangzhou.aliyuncs.com/easyanimate/package/flash_attn-2.6.3%2Bcu118torch2.4cxx11abiFALSE-cp310-cp310-linux_x86_64.whl
|
|
RUN pip install flash_attn-2.6.3+cu118torch2.4cxx11abiFALSE-cp310-cp310-linux_x86_64.whl
|
|
|
|
# install llm-awq (video-caption)
|
|
RUN git clone https://github.com/mit-han-lab/llm-awq /root/llm-awq
|
|
WORKDIR /root/llm-awq
|
|
RUN pip install -e .
|
|
WORKDIR /root/llm-awq/awq/kernels
|
|
# https://github.com/mit-han-lab/llm-awq/issues/93#issuecomment-2144434686
|
|
RUN export TORCH_CUDA_ARCH_LIST="8.0 8.6 8.9 9.0+PTX" && python3 setup.py install
|
|
|
|
# install vllm (video-caption)
|
|
RUN pip install https://github.com/vllm-project/vllm/releases/download/v0.5.4/vllm-0.5.4+cu118-cp310-cp310-manylinux1_x86_64.whl --extra-index-url https://download.pytorch.org/whl/cu118
|
|
|
|
# install requirements (video-caption)
|
|
WORKDIR /root/
|
|
COPY easyanimate/video_caption/requirements.txt /root/requirements-video_caption.txt
|
|
RUN pip install -r /root/requirements-video_caption.txt
|
|
RUN rm /root/requirements-video_caption.txt
|
|
|
|
RUN apt-get update && apt-get install -y aria2
|
|
RUN pip install -U http://eas-data.oss-cn-shanghai.aliyuncs.com/sdk/allspark-0.15-py2.py3-none-any.whl
|
|
RUN pip install -e git+https://github.com/CompVis/taming-transformers.git@master#egg=taming-transformers
|
|
RUN pip install came-pytorch deepspeed pytorch_lightning==1.9.4 func_timeout -i https://mirrors.aliyun.com/pypi/simple/
|
|
|
|
# install requirements
|
|
RUN pip install bitsandbytes mamba-ssm causal-conv1d>=1.4.0 -i https://mirrors.aliyun.com/pypi/simple/
|
|
RUN pip install ipykernel -i https://mirrors.aliyun.com/pypi/simple/
|
|
COPY ./requirements.txt /root/requirements.txt
|
|
RUN pip install -r /root/requirements.txt -i https://mirrors.aliyun.com/pypi/simple/
|
|
RUN rm -rf /root/requirements.txt
|
|
|
|
# install package patches (video-caption)
|
|
COPY easyanimate/video_caption/package_patches/easyocr_detection_patched.py /usr/local/lib/python3.10/dist-packages/easyocr/detection.py
|
|
COPY easyanimate/video_caption/package_patches/vila_siglip_encoder_patched.py /usr/local/lib/python3.10/dist-packages/llava/model/multimodal_encoder/siglip_encoder.py
|
|
|
|
# Use the binary files provided by imageio-ffmpeg without installing the FFmpeg package (video-caption).
|
|
RUN ln -s /usr/local/lib/python3.10/dist-packages/imageio_ffmpeg/binaries/ffmpeg-linux64-v4.2.2 /usr/local/bin/ffmpeg
|
|
|
|
ENV PYTHONUNBUFFERED 1
|
|
ENV NVIDIA_DISABLE_REQUIRE 1
|
|
|
|
WORKDIR /root/ |