alexandre.pires
ec740f458f
- Introduced SMTP settings for Vaultwarden including host, port, security, and authentication details. - Added variables for signup verification, 2FA settings, password hints, and logging options. - Updated Vaultwarden deployment to utilize new SMTP configurations. - Enhanced Grafana module to support dynamic dashboard and datasource provisioning. - Added LLM proxy configuration for Open Web UI with necessary environment variables.
63 lines
1.7 KiB
YAML
63 lines
1.7 KiB
YAML
- name: Setup Qwen3.6 - Coder
|
|
hosts: qwen3_6_host
|
|
vars:
|
|
podman_user: "{{ qwen3_Z6_user | default('qwen36') }}"
|
|
podman_group: "{{ qwen3_6_group | default('qwen36') }}"
|
|
podman_extra_groups: "users,video"
|
|
podman_user_home: "{{ qwen3_6_home | default('/home/qwen36') }}"
|
|
llama_model: "{{ qwen3_6_model | default('bartowski/Qwen_Qwen3.6-35B-A3B-GGUF:IQ4_XS') }}"
|
|
__qwen3_6_context_length__: "{{ qwen3_6_context_length | default(262144) }}"
|
|
podman_user_folders:
|
|
- models
|
|
- .cache
|
|
- .cache/llama.cpp
|
|
pods:
|
|
- name: qwen3.6
|
|
repo:
|
|
image: ghcr.io/ggml-org/llama.cpp:server-vulkan
|
|
ports:
|
|
- "0.0.0.0:{{ qwen3_6_port | default(8012) }}:8080/tcp"
|
|
command:
|
|
- "-t"
|
|
- "12"
|
|
- "-hf"
|
|
- "{{ llama_model }}"
|
|
- "-ngl"
|
|
- "999"
|
|
- "-c"
|
|
- "{{ __qwen3_6_context_length__ }}"
|
|
- "-np"
|
|
- "1"
|
|
- "-b"
|
|
- "1024"
|
|
- "-ub"
|
|
- "512"
|
|
- "-fa"
|
|
- "on"
|
|
- "-ctk"
|
|
- "q4_0"
|
|
- "-ctv"
|
|
- "q8_0"
|
|
- "-cram"
|
|
- "{{ __qwen3_6_context_length__ }}"
|
|
- "-sm"
|
|
- "layer"
|
|
- "-dev"
|
|
- "Vulkan1,Vulkan2"
|
|
- "-ts"
|
|
- "8,12"
|
|
- "-fit"
|
|
- "off"
|
|
- "-mg"
|
|
- 1
|
|
- "--mmap"
|
|
volumes:
|
|
- "{{ podman_user_home }}/models:/models:Z"
|
|
- "{{ podman_user_home }}/.cache:/app/.cache:Z"
|
|
device:
|
|
- "nvidia.com/gpu=all"
|
|
- "/dev/kfd:/dev/kfd:rw"
|
|
- "/dev/dri:/dev/dri:rw"
|
|
roles:
|
|
- podman
|