nm-testing
/

whisper-large-v3-quantized.w8a8_sq

8-bit precision

compressed-tensors

Model card Files Files and versions

whisper-large-v3-quantized.w8a8_sq / recipe.yaml

shubhrapandit's picture

Upload correct model files

03b1e1e about 1 year ago

history blame contribute delete

267 Bytes

	DEFAULT_stage:
	DEFAULT_modifiers:
	SmoothQuantModifier:
	smoothing_strength: 0.9
	ignore: [lm_head]
	GPTQModifier:
	sequential_targets: [WhisperEncoderLayer, WhisperDecoderLayer]
	scheme: W8A8
	targets: Linear
	ignore: [lm_head]