services/homelab
1
1
Fork 0
Code Issues 5 Pull Requests Actions Packages Projects Releases Wiki Activity
Files
ed65f8924ded251cfa4e1c1c496b0a930ce80e95
homelab/active/device_framework_desktop/framework_desktop.md
ducoterra ed65f8924d
All checks were successful
Podman DDNS Image / build-and-push-ddns (push) Successful in 1m13s
Details
add initial framework desktop config
2026-01-13 13:53:52 -05:00

7.3 KiB
Raw Blame History

Framework Desktop

BIOS

https://knowledgebase.frame.work/en_us/changing-memory-allocation-amd-ryzen-ai-max-300-series-By1LG5Yrll

  1. Set GPU memory to 96GB
  2. Add the following kernel args to /etc/default/grub (set allocatable VRAM to 112GB): ttm.pages_limit=28027343750 ttm.page_pool_size=28027343750
  3. Update grub grub2-mkconfig -o /boot/grub2/grub.cfg
  4. Reboot

References

https://docs.podman.io/en/latest/markdown/podman-systemd.unit.5.html

Notes

Update quadlets

scp -r active/device_framework_desktop/quadlets/* deskwork-ai:quadlets/
podman quadlet install --replace quadlets/*

Volume Locations

~/.local/share/containers/storage/volumes/

User

# Create your local ai user. This will be the user you launch podman processes from.
useradd -m ai
loginctl enable-linger ai
su -l ai
mkdir -p ~/.config/containers/systemd/

Llama.cpp

# Build the llama.cpp container image
git clone https://github.com/ggml-org/llama.cpp.git
export BUILD_TAG=$(date +"%Y-%m-%d-%H-%M-%S")
podman build -t llama-cpp-vulkan:${BUILD_TAG} -f .devops/vulkan.Dockerfile .

# Run llama server with gpt-oss-120b
podman run \
-d \
--replace \
--restart always \
--name=llama-server \
-p 8000:8000 \
--device=/dev/kfd \
--device=/dev/dri \
-v llama-server-cache:/root/.cache \
localhost/llama-cpp-vulkan:2026-01-12-10-13-30 \
-hf ggml-org/gpt-oss-120b-GGUF  --ctx-size 32000 --jinja -ub 2048 -b 2048 \
--port 8000 --host 0.0.0.0 -n -1 --n-gpu-layers 999

# To enable autostart, you'll need to create a quadlet
# Quadlets are documented in podman manual pages
# Search for "EXAMPLES" when you run the below command
# Put your quadlet at ~/.config/containers/systemd/
man "podman-systemd.unit(5)"

# Run llama server with devstral-small-2 24b
podman run \
-d \
--name=llama-server-devstral \
--network=host \
--device=/dev/kfd \
--device=/dev/dri \
-v llama-server-cache:/root/.cache \
llama-cpp-vulkan:${BUILD_TAG} \
-hf bartowski/mistralai_Devstral-Small-2-24B-Instruct-2512-GGUF \
--ctx-size 0 --jinja -ub 2048 -b 2048 \
--port 8001 --host 0.0.0.0 -n -1 --n-gpu-layers 999

# Firewall
firewall-cmd --add-port=8000/tcp --permanent
firewall-cmd --reload

Ollama

# Install CLI
curl -fsSL https://ollama.com/download/ollama-linux-amd64.tgz | tar xz -C ~/.local

# Add export OLLAMA_HOST=127.0.0.1
vim ~/.bashrc.d/ollama.sh

# Run ollama
# Will be available on port 11434
podman run \
-d \
--restart always \
--device /dev/kfd \
--device /dev/dri \
-v ollama:/root/.ollama \
-e OLLAMA_VULKAN=1 \
--name ollama \
--network host \
docker.io/ollama/ollama:0.13.5

# Run an image
podman exec -it ollama ollama run gpt-oss:20b

# Firewall
firewall-cmd --add-port=11434/tcp --permanent
firewall-cmd --reload

Anything LLM

Per the docs:

Note --cap-add SYS_ADMIN is a required command if you want to scrape webpages. We use PuppeeteerJS to scrape websites links and --cap-add SYS_ADMIN lets us use sandboxed Chromium across all runtimes for best security practices

mkdir /etc/anything-llm
touch /etc/anything-llm/.env
chown 1000:1000 /etc/anything-llm/.env
chmod 600 /etc/anything-llm/.env

# Add JWT_SECRET=<random string> to this file
vim /etc/anything-llm/.env

# Server will be accessible on port 3001
# Connect llama.cpp as a generic OpenAI LLM provider and use host
# http://172.17.0.1:3001/v1
# Chat model name doesn't matter.
podman run \
-d \
--restart always \
--network host \
--name anythingllm \
--cap-add SYS_ADMIN \
-v anythingllm:/app/server/storage \
-v /etc/anything-llm/.env:/app/server/.env \
-e STORAGE_DIR="/app/server/storage" \
docker.io/mintplexlabs/anythingllm

# Firewall
firewall-cmd --add-port=3001/tcp --permanent
firewall-cmd --reload

Stable Diffusion CPP

# z-turbo
podman run --rm \
-v /home/ai/stable-diffusion.cpp/models:/models:z \
-v /home/ai/stable-diffusion.cpp/output:/output:z \
--device /dev/kfd \
--device /dev/dri \
ghcr.io/leejet/stable-diffusion.cpp:master-vulkan \
--diffusion-model /models/z_turbo/z_image_turbo_bf16.safetensors \
--vae /models/z_turbo/ae.safetensors  \
--llm /models/z_turbo/qwen_3_4b.safetensors \
--cfg-scale 1.0 \
-v \
-H 1024 \
-W 512 \
--seed -1 \
-o /output/output.png \
-p "Framework Laptop 13"

# Flux
podman run --rm \
-v /srv/stable-diffusion.cpp/models:/models:z \
-v ./output:/output:z \
--device /dev/kfd \
--device /dev/dri \
ghcr.io/leejet/stable-diffusion.cpp:master-vulkan \
--diffusion-model  /models/flux/flux1-dev-q4_k.gguf \
--vae /models/flux/ae.safetensors \
--clip_l /models/flux/clip_l.safetensors \
--t5xxl /models/flux/t5xxl_fp16.safetensors \
--cfg-scale 1.0 \
--sampling-method euler \
-v \
-H 512 \
-W 512 \
--seed -1 \
--steps 20 \
-o /output/output.png \
-p "An Everquest video game poster but with ribeye steaks for heads with the words 'EverSteak'"

# Flux2
podman run --rm \
-v /home/ai/stable-diffusion.cpp/models:/models:z \
-v /home/ai/stable-diffusion.cpp/output:/output:z \
--device /dev/kfd \
--device /dev/dri \
ghcr.io/leejet/stable-diffusion.cpp:master-vulkan \
--diffusion-model  /models/flux2/flux2-dev-Q8_0.gguf \
--vae /models/flux2/ae.safetensors \
--llm /models/flux2/Mistral-Small-3.2-24B-Instruct-2506-Q8_0.gguf \
--cfg-scale 1.0 \
--sampling-method euler \
-v \
-H 512 \
-W 1024 \
--seed -1 \
--steps 10 \
-o /output/output.png \
-p "A picture of sign that says 'framework'"

# Qwen
podman run --rm \
-v /home/ai/stable-diffusion.cpp/models:/models:z \
-v /home/ai/stable-diffusion.cpp/output:/output:z \
--device /dev/kfd \
--device /dev/dri \
ghcr.io/leejet/stable-diffusion.cpp:master-vulkan \
--diffusion-model /models/qwen_image/Qwen_Image-Q4_K_M.gguf \
--vae /models/qwen_image/qwen_image_vae.safetensors \
--llm /models/qwen_image/Qwen2.5-VL-7B-Instruct.Q4_K_M.gguf \
--cfg-scale 2.5 \
--sampling-method euler \
-v \
--offload-to-cpu \
-H 512 -W 512 \
--flow-shift 3 \
--seed -1 \
-o /output/output.png \
-p 'Everquest DND mash up poster that says "ever dungeons and dragons"'

# SD3
podman run --rm \
-v /home/ai/stable-diffusion.cpp/models:/models:z \
-v /home/ai/stable-diffusion.cpp/output:/output:z \
--device /dev/kfd \
--device /dev/dri \
ghcr.io/leejet/stable-diffusion.cpp:master-vulkan \
-m  /models/sd3/sd3.5_large.safetensors \
--clip_l /models/sd3/clip_l.safetensors \
--clip_g /models/sd3/clip_g.safetensors \
--t5xxl /models/sd3/t5xxl_fp16.safetensors \
-H 512 -W 512 \
--cfg-scale 4.5 \
--sampling-method euler \
-v \
--seed -1 \
-o /output/output.png \
-p 'Everquest DND mash up poster that says "ever dungeons and dragons"'

Stable Diffusion CPP Server

Uses OpenAI Compatible Endpoints

# z-turbo server
podman run \
-d \
--name stable-diffusion-cpp-server \
-v /srv/stable-diffusion.cpp/models:/models \
-v /srv/stable-diffusion.cpp/build:/output \
--device /dev/kfd \
--device /dev/dri \
--entrypoint "/sd-server" \
--network host \
ghcr.io/leejet/stable-diffusion.cpp:master-vulkan \
--diffusion-model /models/z_turbo/z_image_turbo_bf16.safetensors \
--vae /models/z_turbo/ae.safetensors  \
--llm /models/z_turbo/qwen_3_4b.safetensors \
--cfg-scale 1.0 \
-v \
--diffusion-fa \
-H 1024 \
-W 512 \
--seed -1 \
-l 0.0.0.0

Openai API Web UI

# Will be available on port 8080
podman run \
-d \
--network host \
-v open-webui:/app/backend/data \
--name open-webui \
--restart always \
ghcr.io/open-webui/open-webui:main
Reference in New Issue View Git Blame Copy Permalink