#weight-only • Tags • Transformer System

M8.4 - Scalar Weight-Only PTQ: RTN, GPTQ, and AWQ

🗓 2026-07-14 • Transformer System • ⏱ 42 min read

A derivation- and implementation-oriented treatment of scalar weight-only post-training quantization: round-to-nearest, layer reconstruction geometry, OBD/OBS/OBQ, GPTQ compensation and lazy block updates, AWQ activation-aware scaling, calibration state, packed artifacts, and runtime kernels.

#numinens #transformer #quantization #weight-only #ptq #rtn

Read →

M8.5 - Extreme Weight Compression and Representations

🗓 2026-07-14 • Transformer System • ⏱ 57 min read

A derivation- and implementation-oriented treatment of extreme weight-only compression: sensitivity-weighted nonuniform codebooks, dense-and-sparse exceptions, SpQR's hierarchical representation, AQLM additive codebooks, QuIP# incoherence and E8 lattice codes, VPTQ second-order vector quantization, artifact bit ledgers, and runtime decoder contracts.

#numinens #transformer #quantization #extreme-compression #weight-only #squeezellm

Read →

Su	Mo	Tu	We	Th	Fr	Sa
27	28	29	30	1	2	3
4	5	6	7	8	9	10
11	12	13	14	15	16	17
18	19	20	21	22	23	24
25	26	27	28	29	30	31

| #weight-only

Tag: #weight-only

M8.4 - Scalar Weight-Only PTQ: RTN, GPTQ, and AWQ

M8.5 - Extreme Weight Compression and Representations