nm-testing
/

DeepSeek-R1-Distill-Qwen-14B2of4-sparse

compressed-tensors

Model card Files Files and versions

DeepSeek-R1-Distill-Qwen-14B2of4-sparse / recipe.yaml

horheynm's picture

Upload folder using huggingface_hub

cc32bc2 verified about 1 year ago

177 Bytes

	DEFAULT_stage:
	DEFAULT_modifiers:
	SparseGPTModifier:
	sparsity: 0.5
	mask_structure: '2:4'
	sequential_update: true
	targets: ['re:model.layers.\d*$']