Add new model files and update Ansible playbooks for Qwen3.5 and Qwen3.6
- Created model files for Qwen3.5 with different context lengths. - Added a new Ansible host variable file for CI server configuration. - Updated GPU server host variables to include Ollama models and model files. - Enhanced playbooks for Podman to manage Ollama models, including copying model files and checking existing models. - Introduced a new playbook for setting up Qwen3.6 with specific configurations. - Updated Windows SSH tasks to improve security and configuration management. - Added a Python utility to parse Ollama model names from command output. - Modified Terraform configurations to include new DNS entries and proxy settings. - Improved Nginx proxy configurations with timeout settings.
This commit is contained in:
@@ -0,0 +1,52 @@
|
||||
- name: Setup Qwen3.6 - Coder (FIM - LLAMA.cpp)
|
||||
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(65537) }}"
|
||||
podman_user_folders:
|
||||
- models
|
||||
- .cache
|
||||
- .cache/llama.cpp
|
||||
pods:
|
||||
- name: qwen3.6
|
||||
repo:
|
||||
image: ghcr.io/ggml-org/llama.cpp:server-cuda
|
||||
ports:
|
||||
- "0.0.0.0:{{ qwen3_6_port | default(8012) }}:8080/tcp"
|
||||
command:
|
||||
- "-hf"
|
||||
- "{{ llama_model }}"
|
||||
- "-ngl"
|
||||
- "40"
|
||||
- "-ncmoe"
|
||||
- "32"
|
||||
- "-c"
|
||||
- "{{ __qwen3_6_context_length__ }}"
|
||||
- "-np"
|
||||
- "1"
|
||||
- "-ub"
|
||||
- "512"
|
||||
- "-b"
|
||||
- "1024"
|
||||
- "-fa"
|
||||
- "on"
|
||||
- "--no-mmap"
|
||||
- "-ctk"
|
||||
- "iq4_nl"
|
||||
- "-ctv"
|
||||
- "iq4_nl"
|
||||
- "-cram"
|
||||
- "{{ __qwen3_6_context_length__ }}"
|
||||
volumes:
|
||||
- "{{ podman_user_home }}/models:/models:Z"
|
||||
- "{{ podman_user_home }}/.cache:/app/.cache:Z"
|
||||
device:
|
||||
- "nvidia.com/gpu=all"
|
||||
tasks:
|
||||
- name: Setup Pods
|
||||
ansible.builtin.include_tasks:
|
||||
file: tasks/podman.yml
|
||||
Reference in New Issue
Block a user