janhq
/

Jan-v2-VL-max-FP8

Image-Text-to-Text

compressed-tensors

Model card Files Files and versions

Jan-v2-VL-max-FP8 / recipe.yaml

bachvudinh's picture

Upload folder using huggingface_hub

378632b verified 3 months ago

history blame contribute delete

246 Bytes

	default_stage:
	default_modifiers:
	QuantizationModifier:
	targets: [Linear]
	ignore: [lm_head, 're:.lm_head', 're:visual.', 're:model.visual.', 're:.mlp.gate$',
	're:.*mlp.shared_expert_gate$']
	scheme: FP8_DYNAMIC