Tags

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

b1406

d1031cf4 · sampling : refactor init to use llama_sampling_params (#3696) · Oct 20, 2023
b1405

8cf19d60 · gguf : support big endian platform (#3552) · Oct 20, 2023
b1404

a0edf73b · server : fix uninitialized sampling context (close #3685) · Oct 20, 2023
b1403

f439e506 · ggml : fix rope + llama minor optimizations (#3560) · Oct 20, 2023
b1400

60abea97 · llava : avoid segfault in case of non-existent mmproj file (#3674) · Oct 19, 2023
b1398

4e82b2ea · speculative : bug fixes · Oct 18, 2023
b1397

0e89203b · speculative : add tree-based sampling example (#3624) · Oct 18, 2023
b1396

c67fe68e · metal : implement q5_0 and q5_1 kernels (#3648) · Oct 18, 2023
b1395

1117d066 · opencl : fix element-wise multiplication (#3656) · Oct 18, 2023
b1394

cb33f43a · fix embeddings when using CUDA (#3657) · Oct 17, 2023
b1393

e1675d13 · llama : avoid fprintf in favor of LLAMA_LOG (#3538) · Oct 17, 2023
b1391

40e5ce05 · CLBlast: Fix temporary buffer size for f16 conversion (wsize) · Oct 17, 2023
b1390

a5e8c1d8 · train-text-from-scratch : fix assert failure in ggml-alloc (#3618) · Oct 17, 2023
b1387

1142013d · save-load-state : fix example + add ci test (#3655) · Oct 17, 2023
b1385

1a159553 · tokenizer : special token handling (#3538) · Oct 17, 2023
b1384

281ef73c · k-quants : fix quantization ranges (#3646) · Oct 17, 2023
b1383

940efa95 · llava : fix tokenization to not add bos between image embeddings and user prompt (#3645) · Oct 16, 2023
b1382

11bff290 · MPT : support GQA for replit-code-v1.5 (#3627) · Oct 15, 2023
b1381

11dc1091 · Honor -ngl option for Cuda offloading in llava (#3621) · Oct 14, 2023
b1380

2a4bcbac · llama : remove n_threads from llama_decode_internal (#3614) · Oct 13, 2023

Prev
1
…
50
51
52
53
54
55
56
57
58
…
99
Next