Skip to content

Get up and running with Kimi, GLM, MiniMax, DeepSeek, gpt-oss, Qwen, Gemma and other models.

Commits on main

refactor rope
Michael Yang committed 1 year ago
603ceef
CI: use vendor base commit in cache keys (#13348)
Daniel Hiltgen committed 11 months ago
5dae738
5a41d69
ggml: handle all streams (#13350)
Daniel Hiltgen committed 11 months ago
c146a13
llm: Enable flash attention for mistral3 by default
Jesse Gross committed 11 months ago
9191dfa
ggml: Enable flash attention for vision encoders
Jesse Gross committed 11 months ago
1108d8b
ggml: Always set cache padding to 256
Jesse Gross committed 11 months ago
7837a5b
convert: add deepseek converter (#12980)
Patrick Devine committed 11 months ago
0a844f8
a03223b
ggml update to b7108 (#12992)
Daniel Hiltgen committed 11 months ago
0cf7794
ci: restore previous linter rules (#13322)
Jeffrey Morgan committed 11 months ago
854d40e
app: relay thinking false to server (#13319)
Bruce MacDonald committed 11 months ago
84a2ced
CUDA: filter devices on secondary discovery (#13317)
Daniel Hiltgen committed 11 months ago
3f30836
cc9555a
20aee96
test: avoid ministral tools test on low vram (#13302)
Daniel Hiltgen committed 11 months ago
18b5958
llm: Don't always evict models on CPU-only systems
Jesse Gross committed 12 months ago
5317202
test: add ministral-3 (#13300)
Daniel Hiltgen committed 11 months ago
d771043
CUDA: verify CC is supported by target library (#13298)
Daniel Hiltgen committed 11 months ago
f8f1071
model: ministral w/ llama4 scaling (#13292)
Patrick Devine committed 11 months ago
d3e0a0d
win: warn if ggml-base detected in PATH (#13289)
Daniel Hiltgen committed 11 months ago
5541727
api/client: handle non-json streaming errors (#13007)
Bruce MacDonald committed 11 months ago
5b6a8e6
467bbc0
6d9f932
docs: fix output formatting in faq.mdx (#13231)
Ondrej Kokes committed 11 months ago
0c24896
docs: remove deprecated parameters (#13237)
EntropyYue committed 12 months ago
8b1b89a