Tags

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

b2878

1265c670 · Revert "move ndk code to a new library (#6951)" (#7282) · May 14, 2024
b2877

5e31828d · ggml : add RPC backend (#6829) · May 14, 2024
b2876

54160020 · llama : disable pipeline parallelism with nkvo (#7265) · May 14, 2024
b2875

efc8f767 · move ndk code to a new library (#6951) · May 14, 2024
b2874

e0f55618 · Add left recursion check: quit early instead of going into an infinite loop (#7083) · May 14, 2024
b2871

614d3b91 · llama : less KV padding when FA is off (#7257) · May 13, 2024
b2870

30e70334 · llava-cli: fix base64 prompt (#7248) · May 14, 2024
b2868

948f4ec7 · [SYCL] rm wait() (#7233) · May 13, 2024
b2867

9aa67249 · llama : rename jina tokenizers to v2 (#7249) · May 13, 2024
b2865

e586ee42 · change default temperature of OAI compat API from 0 to 1 (#7226) · May 13, 2024
b2864

cbf75894 · [SYCL] Add oneapi runtime dll files to win release package (#7241) · May 13, 2024
b2862

dc685be4 · CUDA: add FP32 FlashAttention vector kernel (#7188) · May 12, 2024
b2861

6f1b6360 · cmake : fix version cmp (#7227) · May 12, 2024
b2860

b228aba9 · remove convert-lora-to-ggml.py (#7204) · May 12, 2024
b2859

7bd4ffb7 · metal : fix warnings (skipme) (#0) · May 11, 2024
b2854

72c177c1 · fix system prompt handling (#7153) · May 11, 2024
b2852

fae9d234 · sync : ggml · May 11, 2024
b2848

98863133 · server : free llama_batch on exit (#7212) · May 11, 2024
b2847

f99e1e45 · llama : lookup word in vocab before doing BPE merges (#7193) · May 11, 2024
b2846

5ae3426b · server: fix reported top tokens for temperature 0 (#7203) · May 11, 2024

Prev
1
…
4
5
6
7
8
9
10
11
12
…
99
Next