 UEXTM.com
|
5f70671856
Introduce backend GUIDs (ggml/743)
|
hai 1 ano |
 Kawrakow
|
7c4263d426
ggml : make i-quants work with super-blocks of 64 (CPU,Metal) (#5760)
|
hai 1 ano |
|
Kawrakow
|
0becb22ac0
IQ4_XS: a 4.25 bpw quantization (#5747)
|
hai 1 ano |
 Engininja2
|
c24a2a6e60
cuda : replace remaining shfl_xor with calls to warp_reduce functions (#5744)
|
hai 1 ano |
|
Kawrakow
|
a33e6a0d2a
Adding IQ2_S and IQ2_M to complete coverage of the 2-3 bit quantization range (#5721)
|
hai 1 ano |
 Johannes Gäßler
|
47bb7b48c7
CUDA: fix DEBUG_CUDA_MALLOC (#5729)
|
hai 1 ano |
 Georgi Gerganov
|
ab336a9d5e
code : normalize enum names (#5697)
|
hai 1 ano |
|
Kawrakow
|
4c4cb30736
IQ3_S: a much better alternative to Q3_K (#5676)
|
hai 1 ano |
|
Georgi Gerganov
|
7e4f339c40
ggml : always define ggml_fp16_t as uint16_t (#5666)
|
hai 1 ano |
|
Kawrakow
|
a14679cc30
IQ4_NL: 4-bit non-linear quants with blocks of 32 (#5590)
|
hai 1 ano |
 slaren
|
40c3a6c1e1
cuda : ignore peer access already enabled errors (#5597)
|
hai 1 ano |
|
Georgi Gerganov
|
d0e3ce51f4
ci : enable -Werror for CUDA builds (#5579)
|
hai 1 ano |
|
slaren
|
3a9cb4ca64
cuda, metal : fix nans in soft_max (#5574)
|
hai 1 ano |
|
Kawrakow
|
bd2d4e393b
1.5 bit quantization (#5453)
|
hai 1 ano |
|
Georgi Gerganov
|
8f1be0d42f
ggml : add ALiBi support for ggml_soft_max_ext (#5488)
|
hai 1 ano |
|
slaren
|
9060a1e9df
cuda : print message when initialization fails (#5512)
|
hai 1 ano |
|
Johannes Gäßler
|
3bdc4cd0f5
CUDA: mul_mat_vec_q tiling, refactor mul mat logic (#5434)
|
%!s(int64=2) %!d(string=hai) anos |
|
Johannes Gäßler
|
8e6a9d2de0
CUDA: more warps for mmvq on NVIDIA (#5394)
|
%!s(int64=2) %!d(string=hai) anos |
|
Johannes Gäßler
|
aa7ab99be2
CUDA: fixed mmvq kernel for bs 2,3,4 and -sm row (#5386)
|
%!s(int64=2) %!d(string=hai) anos |
|
Johannes Gäßler
|
17c97fb062
CUDA: mul_mat_vec_q max. batch size 8 -> 4 (#5370)
|
%!s(int64=2) %!d(string=hai) anos |
|
Johannes Gäßler
|
2c516611f1
CUDA: mul_mat_vec_q for batch sizes > 1 (#5351)
|
%!s(int64=2) %!d(string=hai) anos |
|
slaren
|
8ca511cade
cuda : fix LLAMA_CUDA_F16 (#5262)
|
%!s(int64=2) %!d(string=hai) anos |
 JidongZhang-THU
|
15606309a0
llava : add MobileVLM support (#5132)
|
%!s(int64=2) %!d(string=hai) anos |
|
Georgi Gerganov
|
8f8ddfcfad
sync : ggml (#0)
|
%!s(int64=2) %!d(string=hai) anos |
 John Balis
|
625a699b54
`ggml_cuda_cpy` support for 4d tensors and float16->float32 upcasting (ggml/686)
|
%!s(int64=2) %!d(string=hai) anos |
|
Kawrakow
|
f4d7e54974
SOTA 3-bit quants (#5196)
|
%!s(int64=2) %!d(string=hai) anos |
 0cc4m
|
2307523d32
ggml : add Vulkan backend (#2059)
|
%!s(int64=2) %!d(string=hai) anos |
|
slaren
|
62fead3ea0
cuda : fix tensor size calculation for non-split buffer (#5145)
|
%!s(int64=2) %!d(string=hai) anos |
|
Engininja2
|
cd4fddb29f
cuda : fix 2-bit quants on amd hip (#5105)
|
%!s(int64=2) %!d(string=hai) anos |
|
Johannes Gäßler
|
9ecdd12e95
CUDA: more info when no device code (#5088)
|
%!s(int64=2) %!d(string=hai) anos |
