Xsn/llama batch ext 2 - #109

Draft

ngxson wants to merge 113 commits into

xsn/llama_batch_extfrom

xsn/llama_batch_ext_2

Commits on Sep 23, 2026

metal : key the fa-vec tuned table by family instead of SKU (#29075)
forforever73
authored
ee3ecce
jinja : parse unary +/- before variables (#29244)
Si Chen
authored
4e416ee
server: fix token counting API crash on sleep (#29309)
Will
authored
42916d8
CUDA: enable sparse-fa for dsv4 prefill (again) (#29298)

am17an
and
ServeurpersoCom
authored
dc9879c
metal: add the missing f32 x bf16 mul_mv variants (#28741)
ServeurpersoCom
authored
9575389
common : keep HF cache dir as path, expose UTF-8 only for logs (#29320)
angt
authored
bddf826
CUDA: add a reserve to avoid spurious warning on older GCC builds (#29317)
am17an
authored
66fba63
ggml : bump version to 0.25.1 (ggml/1637)
ggerganov
committed
e4e2f62
sync : ggml
ggerganov
committed
177cd8c
llama.cpp : bump version to 0.5.0 (#29333)
ggerganov
authored
7fe450e
opencl: add A8 Q6_K non-MoE dp4a binary kernel (#29057)
shaofeiqi
authored
fee39dd
ci : use hf-jobs-cpu-xl runner in server sanitize workflow (#29297)
ggerganov
authored
6e60f35
tests: add `-b/--backend` option to test-llama-archs for testing a specific backend (#27372)

yomaytk
and
JohannesGaessler
authored
d2e5458
server: allow preset to set log file (#29334)
ngxson
authored
b9ae43a
convert : allow vision target for DFlash/Dspark (#29339)
tdakhran
authored
bd4f514

Commits on Sep 24, 2026

scripts : make-release-desc - link previous release in changelog title (#29336)
ggerganov
authored
013b31c
hexagon: reject MUL_MAT_ID when src1 precision is F32 (#29348)
jhen0409
authored
9710a32
test-save-load-state : print a per-model results table in --models mode (#29316)
ggerganov
authored
4c5957c
server,common : fix the GCC 12 stringop-overread false positive (again) (#29325)
angt
authored
2b70583
model : add Ling 3.0 VL support (#29151)
aetherbird
authored
f830688
cuda : add conv3d with implicit GEMM (#29137)
leejet
authored
53ed051
vulkan: tune KHR cooperative matrix support for Adreno GPUs (#29328)
Raman-Raje
authored
3423f94
vulkan: handle misalignment in conv_2d and conv_3d (#29365)
0cc4m
authored
6b790a9
vulkan: int8 coopmat1 matmul implementation for AMD RDNA3 and RDNA4 (#27952)

authored
70c4e15
ci : use hf-jobs-cpu-performance, disable pytest workers (#29369)
ggerganov
authored
70596c4
server : change default pytest workers to 4 (#29376)
danbev
authored
308883b
llama: add llama_batch_ext (#24669)
ngxson
authored
fc343a8
ui : fix missing svg use and animation elements in preview and download (#28962)
nandan2003
authored
945064f
adapt common
ngxson
committed
d954d62
add common_batch
ngxson
committed
38b68c5
wip
ngxson
committed
9ce186d
test: flush status (#28352)
kurquhar
authored
8212c78
wip: spec
ngxson
committed
9c8978b
cont
ngxson
committed
70e4407
common_speculative_process
ngxson
committed
abb6af2
server_batch to use common_batch
ngxson
committed
7c0fca3
rm some stale calls
ngxson
committed
8ce1b95
migrate mtmd
ngxson
committed
c10598d
cuda : add F16 kernel support for CONV_2D_DW (#29064)
crion99
authored
a72e04a
hexagon: support I32 CPY and CONT (#29379)
kurquhar
authored
97a418b
hexagon: dynamic quantizer improvements (#29395)

authored
07fc586
llama-grammar: fix numeric truncation for token_id parsing (#29382)
Daniel Kuts
authored
5cf3a35
hexagon: handle multi-sequence in concat_2d (#29344)
jhen0409
authored
a02c7f5
sync : ggml (#29396)
ggerganov
authored
bced459
metal : optimize sparse FA + clean-up (#29377)
ggerganov
authored
cdc0642
metal : fix graph capture and handle empty graphs (#29390)
ggerganov
authored
84e76d8
handle imrope, handle return val of add()/add_embd()
ngxson
committed
5f1235f

Commits on Sep 25, 2026

Commits on Sep 26, 2026

hexagon: find software divide calls using binary inspection tool (#29449)
trivikram-reddy1
authored
4e74811
opencl: add A8 Q8_0 non-MoE dp4a binary kernel (#29439)
shaofeiqi
authored
9f70b2c
ggml-cpu: tiled mul_mat for k-quants (#27851)

jbooth
and
ggerganov
authored
d834d44
polished Readme and llama-bench (#28968)

truecoder34
and
plotnikov.v10
authored
965f897
jinja : support noncall test statements with arg (#29443)
CISC
authored
a1de614
llama : fix K/V and recurrent state cleanup after failed restores (#27530)
Chipmunk
authored
08618ff
jinja : fix compile error (#29468)
CISC
authored
86a24a1
jinja : implement sameas test (#29448)
CISC
authored
81bc6b8
Revert "Change max context length for auto-fitting with unified KV (#28849)" (#29437)
gaugarg-nv
authored
2145525
server : fix wake_fd warning on Windows (#29479)
angt
authored
fcb3074
cuda: add F16 input to the FWHT (#29096)
bri-prism
authored
6f856c7
musa: build the docker images from the PH1 MUSA SDK image (#29481)
yeahdongcn
authored
694ec23
cuda: support Nemotron 3 Puzzle state size 96 for ssm scan (#28717)
anavp-nvidia
authored
9588757

Commits on Sep 27, 2026

Commits on Sep 28, 2026