Unverified50% confidenceFactExact time
PQ以M=8、K*=256为例，理论组合数高达256^8约1.8×10^19，而存储仅需8字节
1
Sources
50%
Confidence
Long-term
Relevance
7/17/2026
First Seen
Sources
Related Claims
UnverifiedPQ乘积量化通过切分向量、局部聚类、编码替代，可将向量存储体积压缩约8倍（四维示例中从16字节压缩到2字节）74% similarPartially VerifiedA standard FP32 model occupies 4 bytes per parameter, while INT4 quantization compresses this to 0.5 bytes, theoretically achieving an 8x memory saving.70% similarUnverifiedQuantized KV caches (such as storing KV vectors in FP8 or INT4) can compress the cache size to 1/2 or even 1/4 of the original.64% similarUnverifiedFP16格式下，大模型显存需求的估算公式为：参数量 × 2 = 所需显存（GB），即每个参数以16位（2字节）浮点数存储63% similarUnverified一张完整的8-bit GHASH查表需要256×16=4KB，在GPU每个SM通常仅有48-96KB共享内存的情况下会造成内存压力63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/543991API
curl https://kongchang.com/api/v1/knowledge/claims/543991MCP
get_claim(id=543991)