 Georgi Gerganov
|
5b09797321
ggml : remove old quantization functions (#5942)
|
il y a 1 an |
|
Georgi Gerganov
|
8a3012a4ad
ggml : add ggml-common.h to deduplicate shared code (#5940)
|
il y a 1 an |
 bobqianic
|
e25fb4b18f
ggml : use `uint8x16_t` return type for `ggml_vqtbl1q_u8` (#5894)
|
il y a 1 an |
 Jared Van Bortel
|
bd836944f8
quants : use MM256_SET_M128I consistently to fix gcc 7 build (#5889)
|
il y a 1 an |
|
Georgi Gerganov
|
494c870326
ggml : fix IQ3_S AVX implementation (#5834)
|
il y a 1 an |
 Kawrakow
|
bbde6eb256
ggml : IQ3_S improvements (#5829)
|
il y a 1 an |
|
Kawrakow
|
7c4263d426
ggml : make i-quants work with super-blocks of 64 (CPU,Metal) (#5760)
|
il y a 1 an |
|
Kawrakow
|
cb49e0f8c9
Attempt to fix android build (#5752)
|
il y a 1 an |
|
Kawrakow
|
0becb22ac0
IQ4_XS: a 4.25 bpw quantization (#5747)
|
il y a 1 an |
 Engininja2
|
1f30b7a9f1
ggml-quants : fix avx2 iq1_s vec_dot when compiled with gcc (#5742)
|
il y a 1 an |
|
Kawrakow
|
a33e6a0d2a
Adding IQ2_S and IQ2_M to complete coverage of the 2-3 bit quantization range (#5721)
|
il y a 1 an |
 Radosław Gryta
|
abbabc5e51
ggml-quants : provide ggml_vqtbl1q_u8 for 64bit compatibility (#5711)
|
il y a 1 an |
|
Kawrakow
|
4c4cb30736
IQ3_S: a much better alternative to Q3_K (#5676)
|
il y a 1 an |
|
Georgi Gerganov
|
7e4f339c40
ggml : always define ggml_fp16_t as uint16_t (#5666)
|
il y a 1 an |
|
Georgi Gerganov
|
efd56b1c21
ggml : 32-bit arm compat (whisper/1891)
|
il y a 1 an |
|
Kawrakow
|
a14679cc30
IQ4_NL: 4-bit non-linear quants with blocks of 32 (#5590)
|
il y a 1 an |
|
Georgi Gerganov
|
14278f55d2
ggml : restore vec dot stride arg names (#5453)
|
il y a 1 an |
|
Georgi Gerganov
|
b1de96824b
ci : fix wikitext url + compile warnings (#5569)
|
il y a 1 an |
|
Kawrakow
|
bd2d4e393b
1.5 bit quantization (#5453)
|
il y a 1 an |
|
Kawrakow
|
895407f31b
ggml-quants : fix compiler warnings (shadow variable) (#5472)
|
il y a 2 ans |
|
Georgi Gerganov
|
0f2411f154
ggml : fix compile warnings (unused vars) (#4966)
|
il y a 2 ans |
 snadampal
|
a07d0fee1f
ggml : add mmla kernels for quantized GEMM (#4966)
|
il y a 2 ans |
 Michael Podvitskiy
|
b2f87cb64d
ggml : fix `error C2078: too many initializers` for MSVC ARM64 (#5404)
|
il y a 2 ans |
|
Kawrakow
|
f57fadc009
Slight quantization improvement for Q4_K and Q5_K (#5361)
|
il y a 2 ans |
|
Kawrakow
|
6fdfa2ecc6
iq2_xxs: tune quantization (#5320)
|
il y a 2 ans |
|
Kawrakow
|
8e14e3ddb3
Faster AVX2 dot product for IQ2_XS (#5187)
|
il y a 2 ans |
|
Kawrakow
|
f4d7e54974
SOTA 3-bit quants (#5196)
|
il y a 2 ans |
|
Georgi Gerganov
|
38566680cd
ggml : add IQ2 to test-backend-ops + refactoring (#4990)
|
il y a 2 ans |
|
Kawrakow
|
334a835a1c
ggml : importance matrix support for legacy quants (#4969)
|
il y a 2 ans |
|
Kawrakow
|
467a882fd2
Add ability to use importance matrix for all k-quants (#4930)
|
il y a 2 ans |
