nm-testing
/

DeepSeek-R1-Distill-Qwen-14B-W8A8-Dynamic-Per-Token

8-bit precision

compressed-tensors

Model card Files Files and versions Community

DeepSeek-R1-Distill-Qwen-14B-W8A8-Dynamic-Per-Token / recipe.yaml

horheynm's picture

Upload folder using huggingface_hub

830c83d verified 15 days ago

history blame contribute delete

170 Bytes

	DEFAULT_stage:
	DEFAULT_modifiers:
	SmoothQuantModifier: {smoothing_strength: 0.8}
	GPTQModifier:
	scheme: W8A8
	targets: Linear
	ignore: [lm_head]