Tags

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

master-bac6699

bac66994 · Quantization imrovements for k_quants (#2707) · Aug 22, 2023
master-519c981

519c981f · embedding : evaluate prompt in batches (#2713) · Aug 22, 2023
master-1123f7f

1123f7fb · ggml-cuda : use graph allocator (#2684) · Aug 22, 2023
master-ef3f333

ef3f333d · ggml : sync latest (SAM + SD operators, CUDA alibi) (#2709) · Aug 22, 2023
master-8e4364f

8e4364f2 · llama-bench : minor fixes (#2695) · Aug 22, 2023
master-1e3bc52

1e3bc523 · ggml : support CUDA's half type for aarch64(#1455) (#2670) · Aug 22, 2023
master-226255b

226255b4 · server : fallback to default if client param is null (#2688) · Aug 22, 2023
master-6381d4e

6381d4e1 · gguf : new file format with flexible meta data (beta) (#2398) · Aug 21, 2023
master-cb1c072

cb1c0727 · HellaSwag: split token evaluation into batches if needed (#2681) · Aug 21, 2023
master-9e232f0

9e232f02 · ggml : move all type info to ggml_type_traits (#2663) · Aug 20, 2023
master-5e9ff54

5e9ff54a · More efficient Hellaswag implementation (#2677) · Aug 20, 2023
gguf-28b8c26

28b8c265 · cmpnct_gpt2bpe.hpp : cleanup · Aug 19, 2023
master-097e121

097e121e · llama : add benchmark example (#2626) · Aug 18, 2023
master-e9b12c3

e9b12c33 · perplexity : more meaningful ETA number - 2 decimal points · Aug 18, 2023
master-604b8bd

604b8bdf · Fix unicode in grammars (fixes #2501) (#2553) · Aug 17, 2023
master-10151be

10151bee · server : support for saving templates in browser LocalStorage (#2486) · Aug 18, 2023
master-8dae7ce

8dae7ce6 · Add --cfg-negative-prompt-file option for examples (#2591) · Aug 17, 2023
master-a73ccf1

a73ccf1a · llama : replace (permute + reshape + view_1d) with (view_3d) (#2538) · Aug 17, 2023
master-7cf54e1

7cf54e1f · tests : adds simple llama grammar tests (#2618) · Aug 17, 2023
master-a872a2b

a872a2b2 · ggml-alloc : fix discrepency between measure&eval (#2639) · Aug 17, 2023

Prev
1
…
63
64
65
66
67
68
69
70
71
…
99
Next