 Georgi Gerganov
|
3e5aa8a1c4
ggml : fix labels for GGML_OP_ALIBI
|
пре 2 година |
|
Georgi Gerganov
|
c3ca7a5f05
ggml : fix 32-bit ARM NEON
|
пре 2 година |
|
Georgi Gerganov
|
e8c051611a
ggml : use vzip instead of vuzp for consistency
|
пре 2 година |
|
Georgi Gerganov
|
0b5a935099
ggml : fix visibility and unused warnings
|
пре 2 година |
|
Georgi Gerganov
|
ec728e44d7
ggml : fix #if for f32_f32 mul_mat (CLBlast) (#1229)
|
пре 2 година |
|
Georgi Gerganov
|
214b6a3570
ggml : adjust mul_mat_f16 work memory (#1226)
|
пре 2 година |
 slaren
|
7fc50c051a
cuBLAS: use host pinned memory and dequantize while copying (#1207)
|
пре 2 година |
 Henri Vasserman
|
b1ee8f59b4
cuBLAS: non-contiguous tensor support (#1215)
|
пре 2 година |
 Stephan Walter
|
36d19a603b
Remove Q4_3 which is no better than Q5 (#1218)
|
пре 2 година |
|
Georgi Gerganov
|
55390bcaf2
ggml : sync ggml (ggml_alibi)
|
пре 2 година |
|
Georgi Gerganov
|
11d902364b
ggml : add helper debug printf in soft_max
|
пре 2 година |
 0cc4m
|
7296c961d9
ggml : add CLBlast support (#1164)
|
пре 2 година |
 Yann Follet
|
04aaae1d79
add avx2 for dot_q8_0_q8_0, 2x faster than scalar (#1211)
|
пре 2 година |
|
Stephan Walter
|
0b2da20538
ggml : slightly faster AVX2 implementation for Q5 (#1197)
|
пре 2 година |
|
Georgi Gerganov
|
574406dc7e
ggml : add Q5_0 and Q5_1 quantization (#1187)
|
пре 2 година |
|
Georgi Gerganov
|
7a32fcb3b2
ggml : add Q8_0 quantization format (rename the old one to Q8_1) (ARM NEON) (#1179)
|
пре 2 година |
 unbounded
|
dd0eabc049
ggml : use full range for Q4_0 and Q4_2 quantization (#729)
|
пре 2 година |
 xaedes
|
54bb60e268
ggml : fix bug in ggml_compute_forward_sum_f32 (#1162)
|
пре 2 година |
|
Stephan Walter
|
2ec83428de
Fix build for gcc 8 and test in CI (#1154)
|
пре 2 година |
|
Georgi Gerganov
|
ec9cdb6752
ggml : do not print perf ops that have not been used at all
|
пре 2 година |
|
Georgi Gerganov
|
e4422e299c
ggml : better PERF prints + support "LLAMA_PERF=1 make"
|
пре 2 година |
|
Stephan Walter
|
53c8434398
Improve AVX2 for vec_dot_q4_3_q8_0 (#1138)
|
пре 2 година |
 Yishuo Wang
|
c9e2c26f41
A better `packNibbles` and `mul_sum_i8_pairs_float` implementation using AVX512 (#1119)
|
пре 2 година |
|
Georgi Gerganov
|
0e018fe008
ggml : fix Q4_3 cuBLAS
|
пре 2 година |
|
Stephan Walter
|
c50b628810
Fix CI: ARM NEON, quantization unit tests, editorconfig (#1122)
|
пре 2 година |
|
Georgi Gerganov
|
872c365a91
ggml : fix AVX build + update to new Q8_0 format
|
пре 2 година |
|
Georgi Gerganov
|
955ef9a5d5
ggml : alternative Q4_3 implementation using modified Q8_0 (#1109)
|
пре 2 година |
|
Stephan Walter
|
c5aa5e5777
ggml : AVX2 optimization for vec_dot_q4_3_q8_0 and refactoring (#1099)
|
пре 2 година |
|
slaren
|
50cb666b8a
Improve cuBLAS performance by using a memory pool (#1094)
|
пре 2 година |
 Kawrakow
|
1bfc153e2f
ggml : a faster version for Q4_1 x Q8_0 dot products (#1083)
|
пре 2 година |
