Tags
Tags give the ability to mark specific points in history as being important
master-305eb5a
305eb5af
·
build : fix reference to old llama_util.h
·
Apr 29, 2023
master-334637e
334637e4
·
common : change default parameters to pre-#1126 (#1223)
·
Apr 29, 2023
master-dd7eff5
dd7eff57
·
llama : new sampling algorithms (#1126)
·
Apr 29, 2023
master-7fc50c0
7fc50c05
·
cuBLAS: use host pinned memory and dequantize while copying (#1207)
·
Apr 29, 2023
master-b1ee8f5
b1ee8f59
·
cuBLAS: non-contiguous tensor support (#1215)
·
Apr 29, 2023
master-36d19a6
36d19a60
·
Remove Q4_3 which is no better than Q5 (#1218)
·
Apr 28, 2023
master-55390bc
55390bca
·
ggml : sync ggml (ggml_alibi)
·
Apr 28, 2023
master-1481a9c
1481a9cf
·
llama : add session file format and saved sessions in main (#1169)
·
Apr 28, 2023
master-11d9023
11d90236
·
ggml : add helper debug printf in soft_max
·
Apr 28, 2023
master-7296c96
7296c961
·
ggml : add CLBlast support (#1164)
·
Apr 28, 2023
master-92a6e13
92a6e13a
·
Add Manjaro CUDA include and lib dirs to Makefile (#1212)
·
Apr 28, 2023
master-04aaae1
04aaae1d
·
add avx2 for dot_q8_0_q8_0, 2x faster than scalar (#1211)
·
Apr 28, 2023
master-0b2da20
0b2da205
·
ggml : slightly faster AVX2 implementation for Q5 (#1197)
·
Apr 26, 2023
master-574406d
574406dc
·
ggml : add Q5_0 and Q5_1 quantization (#1187)
·
Apr 26, 2023
master-87a6f84
87a6f846
·
Allow setting the rng seed after initialization. (#1184)
·
Apr 26, 2023
master-859fee6
859fee6d
·
quantize : use `map` to assign quantization type from `string` (#1191)
·
Apr 26, 2023
master-7a32fcb
7a32fcb3
·
ggml : add Q8_0 quantization format (rename the old one to Q8_1) (ARM NEON) (#1179)
·
Apr 25, 2023
master-dd0eabc
dd0eabc0
·
ggml : use full range for Q4_0 and Q4_2 quantization (#729)
·
Apr 25, 2023
master-54bb60e
54bb60e2
·
ggml : fix bug in ggml_compute_forward_sum_f32 (#1162)
·
Apr 24, 2023
master-8a0f867
8a0f8673
·
ggml : export symbols (#1155)
·
Apr 24, 2023
1
…
23
24
25
26
27
28
29
30
31
…
38