ngxson/llama.cpp

Pull Requests Commits
mtmd: some small clean up (#17909)
ngxson committed 12 days ago
Verified c6b2c931
cli: enable jinja by default (#17911)
ngxson committed 12 days ago
Verified 34a6d869
server: add presets (config) when using multiple models (#17859)
ServeurpersoCom committed 12 days ago
Verified f32ca51b
Fix race conditions in threadpool when dealing with dynamic/frequent n_threads changes (#17748)
max-krasnyansky committed 12 days ago
Verified e1f49219
ggml : remove GGML_KQ_MASK_PAD constant (#17910)
ggerganov committed 12 days ago
Verified 4dff236a
cuda : add missing support check for xielu (#17895)
CISC committed 12 days ago
Verified 4df6e859
cli: new CLI experience (#17824)
ngxson committed 12 days ago
Verified 6c213177
model : Qwen3-Next-80B-A3B has 48 layers (#17898)
EZForever committed 12 days ago
Verified b6777218
docs : update opencl ops (#17904)
lhez committed 12 days ago
Verified 2d2e1030
CUDA: fix unpadded strides in MMA FA kernel (#17891)
JohannesGaessler committed 12 days ago
Verified 17f7f4ba
convert: allow using quantized Mistral weight (#17889)
ngxson committed 12 days ago
Verified 9e79b011
fix softmax for iGPU (#17838)
NeoZhangJianyu committed 12 days ago
Verified 2e9eab80
common : add parser for ministral/mistral large 3/devstral 2 (#17713)
aldehir committed 13 days ago
Verified 2fbe3b7b
docs : update cpu and cuda ops (#17890)
CISC committed 13 days ago
Verified 63391852
metal: SSM kernel improvements (#17876)
gabe-l-hart committed 13 days ago
Verified 086a63e3
Add DIAG for CUDA (#17873)
pwilkin committed 13 days ago
Verified b6350926
docs: clarify that CPU support should be first (#17886)
JohannesGaessler committed 13 days ago
Verified 48f47565
ggml : Provide macos-specific backtrace printing to avoid terminal death (#17869)
gabe-l-hart committed 13 days ago
Verified 02e409a5
metal : print node names for debugging (#17882)
ggerganov committed 13 days ago
Verified 6b82eb78
ggml : allow fill node alloc inplace (#17870)
CISC committed 13 days ago
Verified 86a3f0fa
cmake: fix Mach-O current version number (#17877)
Rhys-T committed 13 days ago
Verified 63908b63
model : nit, DeepSeek V1 MoE is 16B and GigaChat is 20B (#12652)
CISC committed 13 days ago
Verified 42b12b56
console: allow using arrow left/right, home/end keys and history mode (#17836)
ngxson committed 13 days ago
Verified 4e842d51
CANN: add support for partial RoPE and Vision mode (#17543)
noemotiovon committed 13 days ago
Verified ca709e42
CUDA: fix FP16 overflow in tile FA kernel (#17875)
JohannesGaessler committed 13 days ago
Verified 0cdce38a
llama : add token matching support to llama-grammar (#17816)
aldehir committed 14 days ago
Verified e39502e7
model : support Rnj-1 (#17811)
philip-essential committed 14 days ago
Verified 1d2a1ab7
graph : use fill instead of scale_bias in grouped expert selection (#17867)
CISC committed 14 days ago
Verified c8554b66
model-conversion : add token ids to prompt token output [no ci] (#17863)
danbev committed 14 days ago
Verified 2fa51c19
server: delegate result_state creation to server_task (#17835)
ngxson committed 14 days ago
Verified 951520dd
Terms Of Use Privacy Policy Refund Policy Impressum