implementation detail · filed under inference & serving
Post-RoPE KV-cache quantization
Quantizes the cache after applying RoPE rather than before it.
Also called quantizing the cache after RoPE.
- source
- 1
- model
- 1
- lab adopt it
- 1
- strongest
- default
How sources treat it
One count per evidence span, weakest treatment to strongest.
default 1
Documented in
Evidence
1 span quoted from the sources, strongest treatment first.
We quantize the cache after RoPE: quantizing before RoPE yields only a marginal accuracy improvement in our experiments and would introduce additional overhead during decoding.
defaultunclearin DeepSeek-V4.1-FlashDeepSeek
Filed alongside
Other methods under inference & serving :: inference quantization.
FP8 KV-cache quantizationFP8FP4 KV-cache quantizationNVFP4 quantizationFour-Over-SixMXFP4 weight quantizationPost-training quantizationQuantizationBlock-scaled INT8 quantization with stochastic roundingFP8 E4M3 quantizationGGUFMax-based scalingMSE-based scalingMXFP8 activation quantizationNVFP4 KV-cache quantizationNVFP4 ModelOpt re-quantizationSSM cache quantizationAWQ INT4 weight quantization (W4A16)BF16 inferenceBlock-wise E4M3 FP8 weight quantizationChannel-wise quantizationDynamic activation scalingEmbedding and KV-cache quantizationEmpirical bits-per-element budget selection