noneUsername
/

Devstral-Small-2505-W8A8-Defective

8-bit precision

compressed-tensors

Model card Files Files and versions

Devstral-Small-2505-W8A8-Defective / recipe.yaml

noneUsername's picture

Upload folder using huggingface_hub

c60068f verified 6 months ago

history blame contribute delete

228 Bytes

	default_stage:
	default_modifiers:
	SmoothQuantModifier: {smoothing_strength: 0.87}
	GPTQModifier:
	sequential_update: true
	dampening_frac: 0.01
	scheme: W8A8
	targets: Linear
	ignore: [lm_head]