fedora-copr
/

granite-4.0-h-tiny-quantized.w8a8

Text Generation

granitemoehybrid

Mixture of Experts

8-bit precision

compressed-tensors

Model card Files Files and versions

granite-4.0-h-tiny-quantized.w8a8 / recipe.yaml

jpodivin's picture

Upload folder using huggingface_hub

184b190 verified about 2 months ago

history blame contribute delete

200 Bytes

	default_stage:
	default_modifiers:
	QuantizationModifier:
	targets: [Linear, GraniteMoeHybridParallelExpertsLinear]
	ignore: [lm_head, 're:.*block_sparse_moe.router']
	scheme: W8A8