nytopop
/

Qwen3-8B-abliterated.w8a8

Text Generation

8-bit precision

compressed-tensors

Model card Files Files and versions Community

Qwen3-8B-abliterated.w8a8 / recipe.yaml

nytopop's picture

Upload folder using huggingface_hub

f9580de verified 1 day ago

history blame contribute delete

215 Bytes

	default_stage:
	default_modifiers:
	SmoothQuantModifier: {smoothing_strength: 0.7}
	GPTQModifier:
	dampening_frac: 0.05
	scheme: W8A8
	targets: Linear
	ignore: [lm_head, 're:.*mlp.gate$']