nm-testing
/

fp8_dynamic_moe-e2e

compressed-tensors

Model card Files Files and versions

fp8_dynamic_moe-e2e / recipe.yaml

nm-autobot's picture

Upload folder using huggingface_hub

e4dac70 verified 16 days ago

History Blame Contribute Delete

198 Bytes

	default_stage:
	default_modifiers:
	QuantizationModifier:
	targets: [Linear]
	ignore: [lm_head, 're:.mlp.gate[.].']
	scheme: FP8_DYNAMIC
	bypass_divisibility_checks: false