Files
llm/plugins/model-bundle.yaml

43 lines
1.5 KiB
YAML

id: local-llm-plugin-model-bundle-v1
created_at: 2026-06-19
selection_rule: "One practical high-quality model per plugin, avoiding very large 72B/235B models because GPU VRAM is not yet confirmed."
plugins:
text:
model_card: registry/model-cards/qwen3-4b-instruct-2507.yaml
local_dir: models/base/qwen3-4b-instruct-2507
translation:
model_card: registry/model-cards/lmt-60-4b.yaml
local_dir: models/translation/lmt-60-4b
audio:
model_card: registry/model-cards/whisper-large-v3-turbo.yaml
local_dir: models/audio/whisper-large-v3-turbo
video:
model_card: registry/model-cards/qwen2_5-vl-7b-instruct.yaml
local_dir: models/video/qwen2.5-vl-7b-instruct
image:
model_card: registry/model-cards/sdxl-base-1_0.yaml
local_dir: models/image/sdxl-base-1.0
edit_model_card: registry/model-cards/sdxl-inpainting-1_0.yaml
edit_local_dir: models/image/sdxl-inpainting-1.0
allow_patterns:
- README.md
- LICENSE.md
- model_index.json
- scheduler/*.json
- tokenizer/*
- tokenizer_2/*
- text_encoder/*.json
- text_encoder/*fp16.safetensors
- text_encoder_2/*.json
- text_encoder_2/*fp16.safetensors
- unet/*.json
- unet/*fp16.safetensors
- vae/*.json
- vae/*fp16.safetensors
1c:
model_card: registry/model-cards/qwen3-coder-30b-a3b-instruct-q4_k_m.yaml
local_dir: models/gguf/1c/qwen3-coder-30b-a3b-instruct-q4_k_m
allow_patterns:
- Qwen3-Coder-30B-A3B-Instruct-Q4_K_M.gguf
- README.md