mgoin
/

Mixtral-8x7B-Instruct-v0.1-FP8

Model card Files Files and versions Community

Mixtral-8x7B-Instruct-v0.1-FP8 / recipe.yaml

mgoin's picture

Upload folder using huggingface_hub

cd23aab verified 5 months ago

history blame contribute delete

156 Bytes

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