AI 绘画工具 stable diffusion 安装第三弹

原创于 2025-04-13 15:07:25 发布 · 779 阅读

20 ·

CC 4.0 BY-SA版权

文章标签：

#人工智能 #stable diffusion

人工智能专栏收录该内容

7 篇文章

订阅专栏

Stable-Diffusion-3.5

图片生成

Stable-Diffusion

Stable Diffusion 3.5 (SD 3.5) 是由 Stability AI 推出的新一代文本到图像生成模型，相比 3.0 版本，它提升了图像质量、运行速度和硬件效率

先看看安装部署后的效果，
stable diffusion

stable diffusion webui docker 离线部署
 AI 绘画工具 stable diffusion 离线安装第二弹

在上两篇文章，我们在一台可以联网的机器上，准备好了安装所需要的资源，如模型文件、 apt 包、 pip 包、 git 仓库等，并把它们传送到无法联网的机器上。
其中，下载 git 仓库部分，我们直接克隆了仓库，没有切换到指定的 commit ，在后续使用相应的 git 仓库时，需要注意切换到对应的 commit 。 https://github.com/Stability-AI/stablediffusion.git 下载后的仓库需重命名为 stable-diffusion-stability-ai ，否则会报以下错误，

webui-docker-auto-1  | Installing extension dependencies (if any)
webui-docker-auto-1  | Traceback (most recent call last):
webui-docker-auto-1  |   File "/stable-diffusion-webui/webui.py", line 13, in <module>
webui-docker-auto-1  |     initialize.imports()
webui-docker-auto-1  |   File "/stable-diffusion-webui/modules/initialize.py", line 26, in imports
webui-docker-auto-1  |     from modules import paths, timer, import_hook, errors  # noqa: F401
webui-docker-auto-1  |   File "/stable-diffusion-webui/modules/paths.py", line 34, in <module>
webui-docker-auto-1  |     assert sd_path is not None, f"Couldn't find Stable Diffusion in any of: {possible_sd_paths}"
webui-docker-auto-1  | AssertionError: Couldn't find Stable Diffusion in any of: ['/stable-diffusion-webui/repositories/stable-diffusion-stability-ai', '.', '/']
webui-docker-auto-1 exited with code 1

接下来，我们继续分析 stable-diffusion-webui-docker/services/AUTOMATIC1111 目录下的 Dockerfile 文件内容，原来的 Dockerfile 可看这里 https://github.com/AbdBarho/stable-diffusion-webui-docker/blob/master/services/AUTOMATIC1111/Dockerfile 。

1、基础镜像的使用，
原始 Dockerfile 分为两个阶段：一个用于下载资源的阶段（使用 alpine/git 镜像），一个用于实际应用的阶段（使用 pytorch/pytorch 镜像）。修改后去掉下载资源的阶段，通过离线准备，直接从 pytorch/pytorch:2.3.0-cuda12.1-cudnn8-runtime 镜像开始构建。

# Use PyTorch base image with CUDA
FROM pytorch/pytorch:2.3.0-cuda12.1-cudnn8-runtime as base

2、定义环境变量
设置了一些环境变量，将需要克隆的 Git 仓库的 commit 定义为环境变量后续使用。

# Set environment variables
ENV DEBIAN_FRONTEND=noninteractive PIP_PREFER_BINARY=1

# Define commit hashes as environment variables
ENV STABLE_DIFFUSION_WEBUI_ASSETS_COMMIT=6f7db241d2f8ba7457bac5ca9753331f0c266917
ENV STABLE_DIFFUSION_STABILITY_AI_COMMIT=cf1d67a6fd5ea1aa600c4df58e5b47da45f6bdbf
ENV BLIP_COMMIT=48211a1594f1321b00f14c9f7a5b4813144b2fb9
ENV K_DIFFUSION_COMMIT=ab527a9a6d347f364e3d185ba6d714e22d80cb3c
ENV CLIP_INTERROGATOR_COMMIT=2cf03aaf6e704197fd0dae7c7f96aa59cf1b11c9
ENV GENERATIVE_MODELS_COMMIT=45c443b316737a4ab6e40413d7794a7f5657c19f

ENV GFPGAN_COMMIT=8d2447a2d918f8eba5a4a01463fd48e45126a379
ENV CLIP_COMMIT=d50d76daa670286dd6cacf3bcd80b5e4823fc8e1
ENV OPEN_CLIP_COMMIT=v2.20.0

3、复制本地资源，
将本地准备好的资源复制到镜像中，包括 apt 包、 pip 包， git 仓库。

# Copy all local resources at once
COPY local_resources/apt-packages /local_resources/apt-packages
COPY local_resources/pip-packages /local_resources/pip-packages

4、安装必要的软件包，
通过 dpkg 安装本地的 apt 包，并使用 apt-get -f install 修复依赖关系。

# Install necessary packages
RUN dpkg -i /local_resources/apt-packages/*.deb || true && \
    apt-get -f install -y --no-install-recommends && \
    apt-get clean

5、准备 stable-diffusion-webui，
将本地的 stable-diffusion-webui 资源复制到镜像中，切换到特定的 commit ，安装依赖。

# Create stable-diffusion-webui directory and copy files
COPY local_resources/git-repos/stable-diffusion-webui /stable-diffusion-webui
WORKDIR /stable-diffusion-webui

# Checkout specific commit and install dependencies
RUN git reset --hard v1.9.4 && \
    pip install --no-index --find-links=/local_resources/pip-packages -r requirements_versions.txt

6、准备 Git 仓库资源，
将本地准备好的 Git 仓库资源复制到 repositories 目录，切换到特定的 commit ，

ENV ROOT=/stable-diffusion-webui

# Set up repositories from the copied resources
# stable-diffusion-stability-ai https://github.com/Stability-AI/stablediffusion.git
COPY local_resources/git-repos ${ROOT}/repositories/
COPY local_repos/openai ${ROOT}/openai
RUN cd ${ROOT}/repositories/stable-diffusion-webui-assets && git reset --hard $STABLE_DIFFUSION_WEBUI_ASSETS_COMMIT \
    && cd ${ROOT}/repositories/stable-diffusion-stability-ai && git reset --hard $STABLE_DIFFUSION_STABILITY_AI_COMMIT && rm -rf assets data/**/*.png data/**/*.jpg data/**/*.gif \
    && cd ${ROOT}/repositories/BLIP && git reset --hard $BLIP_COMMIT \
    && cd ${ROOT}/repositories/k-diffusion && git reset --hard $K_DIFFUSION_COMMIT \
    && cd ${ROOT}/repositories/clip-interrogator && git reset --hard $CLIP_INTERROGATOR_COMMIT \
    && cd ${ROOT}/repositories/generative-models && git reset --hard $GENERATIVE_MODELS_COMMIT \
    && mkdir ${ROOT}/interrogate \
    && cp ${ROOT}/repositories/clip-interrogator/clip_interrogator/data/* ${ROOT}/interrogate

7、安装必要的 Python 包，
安装指定版本的 cython 、 pyngrok 、 xformers 、 GFPGAN 、 CLIP 和 open_clip 。

# Install additional Python packages
RUN pip install --no-index --find-links=/local_resources/pip-packages cython pyngrok xformers==0.0.26.post1

# Clone and install specific git repositories with environment variables
RUN git clone ${ROOT}/repositories/GFPGAN /tmp/GFPGAN && \
    cd /tmp/GFPGAN && \
    git checkout ${GFPGAN_COMMIT} && \
    pip install --no-index --find-links=/local_resources/pip-packages .

RUN git clone ${ROOT}/repositories/CLIP /tmp/CLIP && \
    cd /tmp/CLIP && \
    git checkout ${CLIP_COMMIT} && \
    pip install --no-index --find-links=/local_resources/pip-packages .

RUN git clone ${ROOT}/repositories/open_clip /tmp/open_clip && \
    cd /tmp/open_clip && \
    git checkout ${OPEN_CLIP_COMMIT} && \
    pip install --no-index --find-links=/local_resources/pip-packages .

8、安装 libgoogle-perftools 并设置 LD_PRELOAD 。

# Install libgoogle-perftools and set LD_PRELOAD
RUN dpkg -i /local_resources/apt-packages/libgoogle-perftools*.deb || apt-get -f install -y && apt-get clean
ENV LD_PRELOAD=libtcmalloc.so

9、应用的其余部分不修改，

# Copy the rest of the application
COPY . /docker

# Modify gradio/routes.py and set git config
RUN sed -i 's/in_app_dir = .*/in_app_dir = True/g' /opt/conda/lib/python3.10/site-packages/gradio/routes.py && \
    git config --global --add safe.directory '*'

WORKDIR ${ROOT}
ENV NVIDIA_VISIBLE_DEVICES=all
ENV CLI_ARGS=""
EXPOSE 7860

# Set entry point and command
ENTRYPOINT ["/docker/entrypoint.sh"]
CMD python -u webui.py --listen --port 7860 ${CLI_ARGS}

下面是离线部署过程中遇到的一些问题，

问题一：缺少 openai/clip-vit-large-patch14 ，报错信息， OSError: Can't load tokenizer for 'openai/clip-vit-large-patch14'. If you were trying to load it from 'https://huggingface.co/models', make sure you don't have a local directory with the same name. Otherwise, make sure 'openai/clip-vit-large-patch14' is the correct path to a directory containing all relevant files for a CLIPTokenizer tokenizer.
详细日志如下：

webui-docker-auto-1  | creating model quickly: OSError
webui-docker-auto-1  | Traceback (most recent call last):
webui-docker-auto-1  |   File "/opt/conda/lib/python3.10/threading.py", line 973, in _bootstrap
webui-docker-auto-1  |     self._bootstrap_inner()
webui-docker-auto-1  |   File "/opt/conda/lib/python3.10/threading.py", line 1016, in _bootstrap_inner
webui-docker-auto-1  |     self.run()
webui-docker-auto-1  |   File "/opt/conda/lib/python3.10/threading.py", line 953, in run
webui-docker-auto-1  |     self._target(*self._args, **self._kwargs)
webui-docker-auto-1  |   File "/stable-diffusion-webui/modules/initialize.py", line 149, in load_model
webui-docker-auto-1  |     shared.sd_model  # noqa: B018
webui-docker-auto-1  |   File "/stable-diffusion-webui/modules/shared_items.py", line 175, in sd_model
webui-docker-auto-1  |     return modules.sd_models.model_data.get_sd_model()
webui-docker-auto-1  |   File "/stable-diffusion-webui/modules/sd_models.py", line 620, in get_sd_model
webui-docker-auto-1  |     load_model()
webui-docker-auto-1  |   File "/stable-diffusion-webui/modules/sd_models.py", line 723, in load_model
webui-docker-auto-1  |     sd_model = instantiate_from_config(sd_config.model)
webui-docker-auto-1  |   File "/stable-diffusion-webui/repositories/stable-diffusion-stability-ai/ldm/util.py", line 89, in instantiate_from_config
webui-docker-auto-1  |     return get_obj_from_str(config["target"])(**config.get("params", dict()))
webui-docker-auto-1  |   File "/stable-diffusion-webui/repositories/stable-diffusion-stability-ai/ldm/models/diffusion/ddpm.py", line 1650, in __init__
webui-docker-auto-1  |     super().__init__(concat_keys, *args, **kwargs)
webui-docker-auto-1  |   File "/stable-diffusion-webui/repositories/stable-diffusion-stability-ai/ldm/models/diffusion/ddpm.py", line 1515, in __init__
webui-docker-auto-1  |     super().__init__(*args, **kwargs)
webui-docker-auto-1  |   File "/stable-diffusion-webui/repositories/stable-diffusion-stability-ai/ldm/models/diffusion/ddpm.py", line 563, in __init__
webui-docker-auto-1  |     self.instantiate_cond_stage(cond_stage_config)
webui-docker-auto-1  |   File "/stable-diffusion-webui/repositories/stable-diffusion-stability-ai/ldm/models/diffusion/ddpm.py", line 630, in instantiate_cond_stage
webui-docker-auto-1  |     model = instantiate_from_config(config)
webui-docker-auto-1  |   File "/stable-diffusion-webui/repositories/stable-diffusion-stability-ai/ldm/util.py", line 89, in instantiate_from_config
webui-docker-auto-1  |     return get_obj_from_str(config["target"])(**config.get("params", dict()))
webui-docker-auto-1  |   File "/stable-diffusion-webui/repositories/stable-diffusion-stability-ai/ldm/modules/encoders/modules.py", line 103, in __init__
webui-docker-auto-1  |     self.tokenizer = CLIPTokenizer.from_pretrained(version)
webui-docker-auto-1  |   File "/opt/conda/lib/python3.10/site-packages/transformers/tokenization_utils_base.py", line 1809, in from_pretrained
webui-docker-auto-1  |     raise EnvironmentError(
webui-docker-auto-1  | OSError: Can't load tokenizer for 'openai/clip-vit-large-patch14'. If you were trying to load it from 'https://huggingface.co/models', make sure you don't have a local directory with the same name. Otherwise, make sure 'openai/clip-vit-large-patch14' is the correct path to a directory containing all relevant files for a CLIPTokenizer tokenizer.

手工下载 openai/clip-vit-large-patch14 放到本地资源库 local_repos/openai ，同时在 Dockerfile 中增加内容，

COPY local_repos/openai ${ROOT}/openai

问题2，线程创建失败，报错信息如下，

webui-docker-auto-1  | OpenBLAS blas_thread_init: pthread_create failed for thread 63 of 64: Operation not permitted
webui-docker-auto-1  | OpenBLAS blas_thread_init: RLIMIT_NPROC -1 current, -1 max
webui-docker-auto-1  | Traceback (most recent call last):
webui-docker-auto-1  |   File "/stable-diffusion-webui/webui.py", line 13, in <module>
webui-docker-auto-1  |     initialize.imports()
webui-docker-auto-1  |   File "/stable-diffusion-webui/modules/initialize.py", line 17, in imports
webui-docker-auto-1  |     import pytorch_lightning  # noqa: F401
webui-docker-auto-1  |   File "/opt/conda/lib/python3.10/site-packages/pytorch_lightning/__init__.py", line 35, in <module>
webui-docker-auto-1  |     from pytorch_lightning.callbacks import Callback  # noqa: E402
webui-docker-auto-1  |   File "/opt/conda/lib/python3.10/site-packages/pytorch_lightning/callbacks/__init__.py", line 14, in <module>
webui-docker-auto-1  |     from pytorch_lightning.callbacks.batch_size_finder import BatchSizeFinder
webui-docker-auto-1  |   File "/opt/conda/lib/python3.10/site-packages/pytorch_lightning/callbacks/batch_size_finder.py", line 24, in <module>
webui-docker-auto-1  |     from pytorch_lightning.callbacks.callback import Callback
webui-docker-auto-1  |   File "/opt/conda/lib/python3.10/site-packages/pytorch_lightning/callbacks/callback.py", line 25, in <module>
webui-docker-auto-1  |     from pytorch_lightning.utilities.types import STEP_OUTPUT
webui-docker-auto-1  |   File "/opt/conda/lib/python3.10/site-packages/pytorch_lightning/utilities/types.py", line 27, in <module>
webui-docker-auto-1  |     from torchmetrics import Metric
webui-docker-auto-1  |   File "/opt/conda/lib/python3.10/site-packages/torchmetrics/__init__.py", line 26, in <module>
webui-docker-auto-1  |     from torchmetrics import functional  # noqa: E402
webui-docker-auto-1  |   File "/opt/conda/lib/python3.10/site-packages/torchmetrics/functional/__init__.py", line 14, in <module>
webui-docker-auto-1  |     from torchmetrics.functional.audio._deprecated import _permutation_invariant_training as permutation_invariant_training
webui-docker-auto-1  |   File "/opt/conda/lib/python3.10/site-packages/torchmetrics/functional/audio/__init__.py", line 14, in <module>
webui-docker-auto-1  |     from torchmetrics.functional.audio.pit import permutation_invariant_training, pit_permutate
webui-docker-auto-1  |   File "/opt/conda/lib/python3.10/site-packages/torchmetrics/functional/audio/pit.py", line 22, in <module>
webui-docker-auto-1  |     from torchmetrics.utilities import rank_zero_warn
webui-docker-auto-1  |   File "/opt/conda/lib/python3.10/site-packages/torchmetrics/utilities/__init__.py", line 14, in <module>
webui-docker-auto-1  |     from torchmetrics.utilities.checks import check_forward_full_state_property
webui-docker-auto-1  |   File "/opt/conda/lib/python3.10/site-packages/torchmetrics/utilities/checks.py", line 25, in <module>
webui-docker-auto-1  |     from torchmetrics.metric import Metric
webui-docker-auto-1  |   File "/opt/conda/lib/python3.10/site-packages/torchmetrics/metric.py", line 42, in <module>
webui-docker-auto-1  |     from torchmetrics.utilities.plot import _AX_TYPE, _PLOT_OUT_TYPE, plot_single_or_multi_val
webui-docker-auto-1  |   File "/opt/conda/lib/python3.10/site-packages/torchmetrics/utilities/plot.py", line 26, in <module>
webui-docker-auto-1  |     import matplotlib.axes
webui-docker-auto-1  |   File "/opt/conda/lib/python3.10/site-packages/matplotlib/axes/__init__.py", line 1, in <module>
webui-docker-auto-1  |     from . import _base
webui-docker-auto-1  |   File "/opt/conda/lib/python3.10/site-packages/matplotlib/axes/_base.py", line 14, in <module>
webui-docker-auto-1  |     from matplotlib import _api, cbook, _docstring, offsetbox
webui-docker-auto-1  |   File "/opt/conda/lib/python3.10/site-packages/matplotlib/offsetbox.py", line 33, in <module>
webui-docker-auto-1  |     import matplotlib.text as mtext
webui-docker-auto-1  |   File "/opt/conda/lib/python3.10/site-packages/matplotlib/text.py", line 16, in <module>
webui-docker-auto-1  |     from .font_manager import FontProperties
webui-docker-auto-1  |   File "/opt/conda/lib/python3.10/site-packages/matplotlib/font_manager.py", line 1588, in <module>
webui-docker-auto-1  |     fontManager = _load_fontmanager()
webui-docker-auto-1  |   File "/opt/conda/lib/python3.10/site-packages/matplotlib/font_manager.py", line 1582, in _load_fontmanager
webui-docker-auto-1  |     fm = FontManager()
webui-docker-auto-1  |   File "/opt/conda/lib/python3.10/site-packages/matplotlib/font_manager.py", line 1042, in __init__
webui-docker-auto-1  |     timer.start()
webui-docker-auto-1  |   File "/opt/conda/lib/python3.10/threading.py", line 935, in start
webui-docker-auto-1  |     _start_new_thread(self._bootstrap, ())
webui-docker-auto-1  | RuntimeError: can't start new thread
webui-docker-auto-1 exited with code 1