nm-testing
/

Phi-3.5-MoE-instruct-FP8

compressed-tensors

Model card Files Files and versions Community

Phi-3.5-MoE-instruct-FP8 / recipe.yaml

alexmarques's picture

Upload folder using huggingface_hub

5a393d8 verified 2 months ago

history blame contribute delete

152 Bytes

	quant_stage:
	quant_modifiers:
	QuantizationModifier:
	ignore: [lm_head, 're:.*block_sparse_moe.gate']
	scheme: FP8
	targets: Linear