RedHatAI
/

Qwen2.5-VL-72B-Instruct-quantized.w8a8

text-generation-inference

8-bit precision

compressed-tensors

Model card Files Files and versions

Qwen2.5-VL-72B-Instruct-quantized.w8a8 / recipe.yaml

Shubhra Pandit

Upload model files

4699ffe 9 months ago

history blame contribute delete

212 Bytes

	DEFAULT_stage:
	DEFAULT_modifiers:
	GPTQModifier:
	sequential_targets: [Qwen2_5_VLDecoderLayer]
	dampening_frac: 0.03
	scheme: W8A8
	targets: Linear
	ignore: [lm_head, 're:visual.*']