Files
Qwen3-VL-8B-Instruct-W4A16/recipe.yaml

11 lines
285 B
YAML
Raw Permalink Normal View History

default_stage:
default_modifiers:
GPTQModifier:
targets: [Linear]
ignore: ['re:.*lm_head', 're:visual.*', 're:model.visual.*', 're:.*mlp.gate$']
scheme: W4A16
block_size: 128
dampening_frac: 0.01
actorder: static
offload_hessians: false