Tags

Tags give the ability to mark specific points in history as being important

master-ff966e7

ff966e7c · build : fix several cast and printf warnings (#2499) · Aug 04, 2023
master-468ea24

468ea24f · CUDA: faster non k-quant mul_mat_q kernels (#2483) · Aug 02, 2023
master-4f6b60c

4f6b60c7 · CUDA: Fix models with output size != 32000 (#2480) · Aug 02, 2023
master-81844fb

81844fbc · tests : Fix compilation warnings (Linux/GCC) (#2451) · Aug 02, 2023
master-86aeb27

86aeb277 · server : Support dark mode (#2414) · Aug 01, 2023
master-49e7cb5

49e7cb5b · CUDA: fixed LLAMA_FAST compilation option (#2473) · Jul 31, 2023
master-b772bba

b772bba4 · CUDA: fixed cmake F16 option (#2471) · Jul 31, 2023
master-0728c5a

0728c5a8 · CUDA: mmq CLI option, fixed mmq build issues (#2453) · Jul 31, 2023
master-1215ed7

1215ed7d · CUDA: Implemented row flattening for non-glm RoPE (#2468) · Jul 31, 2023
master-2dbf518

2dbf5189 · CUDA: fewer memory bank conflicts for mul_mat_q (#2458) · Jul 31, 2023
master-9d2382b

9d2382b3 · Fix Metal backend broken from the allocator changes (#2455) · Jul 31, 2023
master-a113689

a1136895 · ggml : add graph tensor allocator (#2411) · Jul 30, 2023
master-11f3ca0

11f3ca06 · CUDA: Quantized matrix matrix multiplication (#2160) · Jul 29, 2023
master-9baf9ef

9baf9ef3 · CUDA: faster multi GPU synchronization (#2448) · Jul 29, 2023
master-8a88e58

8a88e585 · perplexity : add Hellaswag calculation (#2389) · Jul 28, 2023
master-a9559bf

a9559bf7 · ggml : workaround for missing _mm256_setr_m128i in GCC < 8 in k_quants.c (#2405) · Jul 28, 2023
master-ee1b497

ee1b497c · llama : support more diverse tokenizers? (#2420) · Jul 28, 2023
master-1a94186

1a941869 · metal : disable graph concurrency optimization due to bug (#2413) · Jul 27, 2023
master-b5472ea

b5472ea0 · ggml : fix assert in ggml_set_unary_op (#2410) · Jul 26, 2023
master-6df1f59

6df1f594 · make : build with -Wmissing-prototypes (#2394) · Jul 26, 2023

Prev
1
…
66
67
68
69
70
71
72
73
74
…
99
Next