mickr777

Sound like something didnt get downloaded correctly, try deleting the pyvenv.cfg file under the appdata/invokeai/venv/ folder and rerun the container

Fixed the start.sh script as something was changed that caused the models.yaml to not be created in right folder

Im not sure I haven't focused on that side of it, but you can ask here https://github.com/invoke-ai/InvokeAI/discussions

Diffusers are meant to more stream lined in the back end, easier to train and seems to be they way everything is going. never trust the example images on any model, as they always pick the best examples to show also look into negative prompting after your normal prompt, add text into [ ] eg. [cropped, worst quality, low quality, artifacts, blurry, bad hands], everything that goes in negative prompt [ ] it tries not to add to image. I found anything v3.0 is better at anime (but best to use a negative prompt [nude] as it can be a little nsfw) anythingv-3.0: description: Anything V3 - a latent diffusion model for anime repo_id: Linaqruf/anything-v3.0 format: diffusers width: 512 height: 512

inpainting is a little more hungry on vram then txt2image I think you need at least 6gb of vram at this moment I haven't had that issue my self but I have the diffuser version of waifu diffusion running you can try the diffuser version by manually adding this to your /userfiles/config/models.yaml, then on first switch to it in the webui it will download the required files waifu-diffusion-v1.4: description: has been conditioned on high-quality anime images through fine-tuning repo_id: hakurei/waifu-diffusion format: diffusers width: 512 height: 512

No problem, it can be called what ever you like anyway, as long as its the same in the xml

that normally happens if the docker image doesn't exist locally. when making the docker image did you use this docker build . -t invokeai_docker as the name there needs to match the repo in the xml, as we are only making a local image not an online one

If anyone gets a crash and docker exits after the diffuser update today, delete the contents of the folder \userfiles\models\hub\ and rerun the docker

thanks you, it looks like when no hugginface token is given, the folders are not created I have updated start.sh and made some notes at bottom of first post

Ah I just saw you purchased a Telsa m40, worse case you might have to create a windows or linux vm and passthrough the gpu and install the gpu driver in the vm and then InvokeAI in it using there installer. (but that is outside the scope of my guide) https://github.com/invoke-ai/InvokeAI/releases/tag/v2.2.5

Did you install the unraid nvidia driver plug in? plus it is a good idea to install NVTOP and GPU Statistics plugins with it too Also in your my-invokeai.xml only change the port like this leave the rest 9090, if your using a different default port <Config Name="Webui Port" Target="9090" Default="9090" Mode="tcp" Description="" Type="Port" Display="always" Required="true" Mask="false">7790</Config>

Looks like their build script makes the invokeai.init file now on start, so I needed to change my script a little, but if you edit and delete everything in the userfiles/invokeai.init file and just add --web --host="0.0.0.0" to it should fix it

Ok I Updated the script, try building again you will need to remove the docker, delete the img and the folders made and start the guide again try <ExtraParams>--gpus 0</ExtraParams>

Yes you can run from cpu, but is extremely slow, to do this in my-invokeai.xml from the guide change <ExtraParams>--gpus all</ExtraParams> to <ExtraParams>--gpus 0</ExtraParams>

If I am not mistaken that gpu only has 2gb vram, You need to have a gpu with at least 4gb vram (8gb is highly recommended)

No related to Invoke AI docker in community app (Still Works as of July 2026) This is a simple unofficial Docker for Unraid (with most data stored in appdata and uses git pull, this helps base image to not need constant updates) for InvokeAI: A Stable Diffusion Toolkit, on start it will check for updates from git, update python venv if changes are needed and auto start the web ui This Docker is using the main branch of InvokeAI, as it is getting updated all the time, things might break every now and then check out their hard work @ https://github.com/invoke-ai/InvokeAI - Changelog or join the InvokeAI discord https://discord.gg/ZmtBAhwWhy https://github.com/invoke-ai/InvokeAI/releases I Know only a small amount about making a docker so this is probably not the right way as I just used google to work it all out, but it worked for me, thought others might find it useful. My Opinion For best Experience (may or may not run on other hardware): Minimal Requirements: A Nvidia GPU with at least 6gb Vram, at least a Pascal based cards, 12gb free system ram, 20gb storage space Recommended requirements: A Nvidia GPU with 8gb+ Vram, Turing based cards+, 16gb+ free system ram, 40gb+ storage space There is now 2 Options to make the docker container: Option 1 - Simplified Install (using Docker Hub) 1. building the docker container make a file called my-invokeai.xml or my-invokeai_prenodes.xml in \config\plugins\dockerMan\templates-user (on your unraid flash drive) add the text below from one of the options below to it and save then go to your unraid gui and then the docker tab, click "add container", in template dropdown box select your user template "invokeai" change port/host paths if required add your hugging face access token if you wish to have auto download of some models/concepts/diffusers <?xml version="1.0"?> <Container version="2"> <Name>InvokeAI</Name> <Repository>mickr777/invokeai_unraid_main</Repository> <Registry>https://hub.docker.com/r/mickr777/invokeai_unraid_main</Registry> <Network>bridge</Network> <MyIP/> <Shell>bash</Shell> <Privileged>false</Privileged> <Support>https://forums.unraid.net/topic/130913-guide-invokeai-a-stable-diffusion-toolkit-docker/</Support> <Project>https://github.com/invoke-ai/InvokeAI/</Project> <Overview>Simplified Docker with auto update for InvokeAI and Unraid.</Overview> <Category>Other: Status:Beta</Category> <WebUI>http://[IP]:[PORT:9090]/</WebUI> <TemplateURL/> <Icon>https://i.ibb.co/N2c008N/invokeai.png</Icon> <ExtraParams>--gpus all</ExtraParams> <PostArgs/> <CPUset/> <DateInstalled/> <DonateText/> <DonateLink/> <Requires/> <Config Name="InvokeAI" Target="/home/invokeuser/InvokeAI/" Default="/mnt/cache/appdata/invokeai/invokeai/" Mode="rw" Description="InvokeAI source code path" Type="Path" Display="always" Required="true" Mask="false">/mnt/cache/appdata/invokeai/invokeai/</Config> <Config Name="userfiles" Target="/home/invokeuser/userfiles/" Default="/mnt/cache/appdata/invokeai/userfiles/" Mode="rw" Description="Persistent storage for models, configs, and outputs" Type="Path" Display="always" Required="true" Mask="false">/mnt/cache/appdata/invokeai/userfiles/</Config> <Config Name="venv" Target="/home/invokeuser/venv/" Default="/mnt/cache/appdata/invokeai/venv/" Mode="rw" Description="Python virtual environment storage" Type="Path" Display="always" Required="true" Mask="false">/mnt/cache/appdata/invokeai/venv/</Config> <Config Name="cache" Target="/home/invokeuser/.cache" Default="/mnt/user/appdata/invokeai/cache" Mode="rw" Description="Cache directory (e.g., Hugging Face models)" Type="Path" Display="always" Required="true" Mask="false">/mnt/user/appdata/invokeai/cache</Config> <Config Name="Webui Port" Target="9090" Default="9090" Mode="tcp" Description="WebUI port" Type="Port" Display="always" Required="true" Mask="false">9090</Config> </Container> 2. Last steps after container build, on first run the python venv will be created and the preload of some models/weights/diffusers (this can take a while and will download 20gb+ of data, open docker log for progress) once this is done load up any web browser and point it to [Your Unraid IP]:9090 (or the port you set) --------------------------------------------------------------------------------------------- Option 2 - Full Manual Install - Main Branch (or just want to see what i did to get it to work) 1. First we will create the needed files create a folder on one of the drive's on you unraid server (I named mine invokeai) inside that folder create a text file called Dockerfile (make sure you remove the txt extension if it has one) add this info in to it and save: FROM ubuntu:24.04 RUN apt-get update \ && DEBIAN_FRONTEND="noninteractive" \ apt-get install -y \ git \ dos2unix \ python3-pip \ python3-venv \ libopencv-dev \ sudo \ && apt-get clean RUN apt-get install -y curl sudo RUN curl -fsSL https://deb.nodesource.com/setup_22.x | sudo -E bash - && sudo apt-get install -y nodejs RUN npm install -g pnpm RUN useradd --create-home -u 99 -g 100 invokeuser && \ chown -R 99:100 /home/invokeuser WORKDIR /home/invokeuser ADD start.sh . RUN dos2unix start.sh RUN chmod +x start.sh USER invokeuser WORKDIR /home/invokeuser ADD invokeai.yaml . ENTRYPOINT ["/bin/bash", "start.sh"] then create a text file called start.sh add this info in to it and click save: #!/bin/bash # Sets a variable for the home directory HOMEDIR="/home/invokeuser" # Checks if the git repo has been cloned and clones it if needed if [ -f "$HOMEDIR/InvokeAI/pyproject.toml" ]; then git config --global --add safe.directory "$HOMEDIR/InvokeAI" cd "$HOMEDIR/InvokeAI" || exit else echo "Cloning Git Repo into Local Folder..." git config --global --add safe.directory "$HOMEDIR/InvokeAI" git clone -b main https://github.com/invoke-ai/InvokeAI.git InvokeAI cd "$HOMEDIR/InvokeAI" || exit fi # Checks if Python Environment has been made and creates it if it has not if [ -f "$HOMEDIR/venv/pyvenv.cfg" ]; then source "$HOMEDIR/venv/bin/activate" else echo "Creating Python Environment...." python3 -m venv "$HOMEDIR/venv/" source "$HOMEDIR/venv/bin/activate" pip install --use-pep517 --no-cache-dir . --extra-index-url https://download.pytorch.org/whl/cu126 fi # Checks if the git repo has had any changes and updates if needed echo "Checking if The Git Repo Has Changed...." git fetch UPSTREAM=${1:-'@{u}'} LOCAL=$(git rev-parse @) REMOTE=$(git rev-parse "$UPSTREAM") BASE=$(git merge-base @ "$UPSTREAM") if [ "$LOCAL" = "$REMOTE" ]; then echo "Local Files Are Up to Date" elif [ "$LOCAL" = "$BASE" ]; then echo "Updates Found, Updating the local Files...." git stash git config pull.rebase true git pull git stash pop fi # Gets date modified for upcoming if statements current=$(date +%s) last_modified_env=$(stat -c "%Y" "$HOMEDIR/InvokeAI/pyproject.toml") # Updates Python Environment if changes have been made to the pyproject.toml file if [ -f "$HOMEDIR/venv/pyvenv.cfg" ] && [ $((current - last_modified_env)) -lt 60 ]; then echo "Updates Found, Updating Python Environment...." pip install --use-pep517 --no-cache-dir --upgrade -e . fi # Check for changes in the frontend and build if necessary cd "$HOMEDIR/InvokeAI/invokeai/frontend/web/" || exit # Get the latest commit hash for the frontend directory FRONTEND_COMMIT=$(git log -n 1 --pretty=format:"%H" -- .) # Path to commit hash file COMMIT_HASH_FILE="$HOMEDIR/userfiles/frontend_commit_hash.txt" FRONTEND_DIR="$HOMEDIR/InvokeAI/invokeai/frontend/web" DIST_DIR="$FRONTEND_DIR/dist" # Get the latest commit hash for the frontend directory FRONTEND_COMMIT=$(git -C "$HOMEDIR/InvokeAI" log -n 1 --pretty=format:"%H" -- invokeai/frontend/web) # Function to build frontend build_frontend() { echo "Running pnpm install and build..." cd "$FRONTEND_DIR" || exit pnpm install pnpm build echo "$FRONTEND_COMMIT" > "$COMMIT_HASH_FILE" cp -r /home/invokeuser/InvokeAI/invokeai/frontend/web/dist \ /home/invokeuser/venv/lib/python3.12/site-packages/invokeai/frontend/web/ } # Check if build is needed if [ ! -f "$COMMIT_HASH_FILE" ]; then echo "No previous frontend build detected. Building the frontend..." build_frontend else PREV_FRONTEND_COMMIT=$(cat "$COMMIT_HASH_FILE") if [ "$FRONTEND_COMMIT" != "$PREV_FRONTEND_COMMIT" ]; then echo "Frontend changes detected. Rebuilding the frontend..." build_frontend elif [ ! -f "$DIST_DIR/index.html" ]; then echo "Build output missing. Rebuilding the frontend..." build_frontend else echo "Frontend is up to date. No rebuild necessary." fi fi CONFIG_FILE="$HOMEDIR/userfiles/invokeai.yaml" LINE_TO_CHECK="host: 0.0.0.0" mkdir -p "$HOMEDIR/userfiles" if [ ! -f "$CONFIG_FILE" ]; then echo "Config file missing. Copying new one..." cp "$HOMEDIR/invokeai.yaml" "$CONFIG_FILE" elif ! grep -qF "$LINE_TO_CHECK" "$CONFIG_FILE"; then echo "Line not found. Overwriting config file..." cp "$HOMEDIR/invokeai.yaml" "$CONFIG_FILE" else echo "Config file already contains the line. No action needed." fi # Start the web UI and the backend echo "Loading InvokeAI WebUI....." cd "$HOMEDIR/InvokeAI" python scripts/invokeai-web.py --root="$HOMEDIR/userfiles/"Create a file called invokeai.yaml # Internal metadata - do not edit: schema_version: 4.0.2 # Put user settings here - see https://invoke-ai.github.io/InvokeAI/configuration/: host: 0.0.0.0 2. Building the docker image Open the Unraid Terminal cd to folder where the 2 files we just created are stored run this command and wait for it to be done: docker build . -t invokeai_docker_main 3. building the docker container make a file called my-invokeai.xml in \config\plugins\dockerMan\templates-user (on your unraid flash drive) add the text below to it and save then go to your unraid gui and then the docker tab, click "add container", in template dropdown box select your user template "invokeai" change port/host paths if required <?xml version="1.0"?> <Container version="2"> <Name>InvokeAI</Name> <Repository>invokeai_docker_main</Repository> <Registry>https://hub.docker.com/r/mickr777/invokeai_unraid_main</Registry> <Network>bridge</Network> <MyIP/> <Shell>bash</Shell> <Privileged>false</Privileged> <Support>https://forums.unraid.net/topic/130913-guide-invokeai-a-stable-diffusion-toolkit-docker/</Support> <Project>https://github.com/invoke-ai/InvokeAI/</Project> <Overview>Simplified Docker with auto update for InvokeAI and Unraid.</Overview> <Category>Other: Status:Beta</Category> <WebUI>http://[IP]:[PORT:9090]/</WebUI> <TemplateURL/> <Icon>https://i.ibb.co/N2c008N/invokeai.png</Icon> <ExtraParams>--gpus all</ExtraParams> <PostArgs/> <CPUset/> <DateInstalled/> <DonateText/> <DonateLink/> <Requires/> <Config Name="InvokeAI" Target="/home/invokeuser/InvokeAI/" Default="/mnt/cache/appdata/invokeai/invokeai/" Mode="rw" Description="InvokeAI source code path" Type="Path" Display="always" Required="true" Mask="false">/mnt/cache/appdata/invokeai/invokeai/</Config> <Config Name="userfiles" Target="/home/invokeuser/userfiles/" Default="/mnt/cache/appdata/invokeai/userfiles/" Mode="rw" Description="Persistent storage for models, configs, and outputs" Type="Path" Display="always" Required="true" Mask="false">/mnt/cache/appdata/invokeai/userfiles/</Config> <Config Name="venv" Target="/home/invokeuser/venv/" Default="/mnt/cache/appdata/invokeai/venv/" Mode="rw" Description="Python virtual environment storage" Type="Path" Display="always" Required="true" Mask="false">/mnt/cache/appdata/invokeai/venv/</Config> <Config Name="cache" Target="/home/invokeuser/.cache" Default="/mnt/user/appdata/invokeai/cache" Mode="rw" Description="Cache directory (e.g., Hugging Face models)" Type="Path" Display="always" Required="true" Mask="false">/mnt/user/appdata/invokeai/cache</Config> <Config Name="Webui Port" Target="9090" Default="9090" Mode="tcp" Description="WebUI port" Type="Port" Display="always" Required="true" Mask="false">9090</Config> </Container> 4. Last steps after container build, on first run the python venv will be created once this is done load up any web browser and point it to [Your Unraid IP]:9090 (or the port you set) Last Notes: if you run in to errors after updates, cleaning out the /invokeai/invokeai/ folder and/or deleting the file /invokeai/venv/pyvenv.cfg and rerunning the docker can force a part rebuild and fix a lot of issues Feel free to comment with any suggestions