Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
606 commits
Select commit Hold shift + click to select a range
0df974d
sycl : enhance the api to support peer-to-peer copy (#27550)
arthw Sep 3, 2026
c7bda03
vulkan: fix FA dequant path engagement (#28190)
Nathanw1014 Sep 3, 2026
0ba6499
CUDA: Allow concurrent streams per split for multi-GPU (#28198)
tannerbruhn Sep 3, 2026
f45576a
mtmd : add `const` in various places (#28307)
madsmtm Sep 3, 2026
0df017d
metal : fix glu dispatch with ne00 = 1 (#28306)
ggerganov Sep 3, 2026
7bb0fc1
metal : add sparse FA (#28098)
ggerganov Sep 3, 2026
de8656b
mtmd: propagate const to preproc class (#28310)
ngxson Sep 3, 2026
42f0225
server : use pytest-xdist for server tests (#28298)
danbev Sep 3, 2026
e107984
ops: add Hexagon to ops.md and update main README.md (#28263)
tboinovski1 Sep 3, 2026
d30500b
snapdragon: ci updates to use new run script (#28293)
max-krasnyansky Sep 3, 2026
8c1a251
tune MMVQ to MMQ crossover for SM87 (#28285)
kbenkhaled Sep 3, 2026
95ef7fc
opencl: quant lm_head / decode GEMV and medium-batch GEMM optimizatio…
wanghqc Sep 3, 2026
c5a5535
common/json-schema : fix GBNF grammar generation for empty object sch…
SergeySklyarov Sep 3, 2026
d230ddd
llama: fix whole source code rebuilt on each new commit (#28278)
ngxson Sep 3, 2026
f9f09f0
SYCL: Refactor GGML_SYCL_ENABLE_MKL_FA to global var (#26863)
johnkarlhill Sep 4, 2026
6703d78
sycl: fuse rms_norm+mul+add and add+add residual chains (#27610)
newjordan Sep 4, 2026
9a4843c
src : add n_expert_used_max function (#28323)
danbev Sep 4, 2026
832fd6f
ggml-cpu(s390x) : fix q5_1 uninitialized v_acc (#28332)
taronaeo Sep 4, 2026
c390d0a
common : make build info output stream configurable (#28322)
angt Sep 4, 2026
0ef4d56
ci : disable failing openvino tests (#28347)
ravi9 Sep 4, 2026
38521ec
vendor: update BoringSSL to 0.20260903.0 (#28354)
cabelo Sep 4, 2026
a529af9
docs : update maintainer PRs link and regenerate AUTHORS (#28365)
ggerganov Sep 4, 2026
24f5bf8
ggml : remove GGML_CUDA_PEER_MAX_BATCH_SIZE (#28177)
angt Sep 4, 2026
4cbe8b0
ggml : don't crash when backend search path can't be read (#28271)
angt Sep 4, 2026
d509cb1
Don't use npx inside a package.json script (#28270)
eapache Sep 4, 2026
86b351f
ggml : replace compile definitions with version.h.in (#28364)
danbev Sep 4, 2026
8f83678
metal : add remaining fa-vec tunings for M3 Max (#28373)
nikwen Sep 4, 2026
163a407
model, mtmd: fix gemma4 vision handling (#28335)
ngxson Sep 4, 2026
64a155d
sync : ggml (#28379)
ggerganov Sep 4, 2026
5266f24
llama.cpp : bump version to 0.4.0 (#28386)
ggerganov Sep 4, 2026
49c0dc8
model : add Tencent Hy 4 (hy_v4) preview architecture support (#28127)
Little0o0 Sep 4, 2026
1863ac0
ui: export conversations from database instead of cached store (#27432)
intel00000 Sep 4, 2026
8b4b355
ci : move more jobs to ccache-buckets (#28375)
CISC Sep 4, 2026
4acf4a4
opencl: add Adreno xmem SDPA path (#26331)
happyyzy Sep 4, 2026
1548a24
opencl: extend the elementwise and data‐movement op coverage (#27633)
wanghqc Sep 4, 2026
85d5703
ui : fix MCP image attachments not displayed in tool block (#25789) (…
nachobh Sep 4, 2026
427291b
metal : add remaining fa-vec tunings for M3 (#28396)
init-22 Sep 4, 2026
cd8cdf3
sycl: attribute device allocations by site (GGML_SYCL_MEMTRACE) (#27631)
nicois Sep 5, 2026
4d91760
sycl : fix test-backend-ops CI break && restore Kronecker product FWH…
philip-jingxin Sep 5, 2026
6a1a922
metal : fix memory leak in early return (#28399)
nikwen Sep 5, 2026
74a7c89
Github: limit blank issues to maintainers (#28435)
JohannesGaessler Sep 5, 2026
971595d
metal : add remaining fa-vec tunings for M2 Max (#28458)
nikwen Sep 6, 2026
c457e3b
ui : embed assets directly with CMake (#28445)
angt Sep 6, 2026
7620399
common: add --log-jsonl (#28437)
ngxson Sep 6, 2026
0afb805
ui: Improve Chat Messages rendering performance (#28460)
allozaur Sep 6, 2026
9e0e220
grammar : fix max repetition threshold (#28469)
aldehir Sep 6, 2026
73a43d1
cuda: fixes races in mmid and mmf (#28475)
am17an Sep 6, 2026
d03efa5
opencl: properly choose weights pack for q4_K, q5_K mul_mat (#28402)
lhez Sep 6, 2026
3ad1ba7
[Model] Support for Spark2_5ForCausalLM implementation (#27868)
KnightYao Sep 6, 2026
5fdfa62
models : fix GDN normalization from `max` to `rsqrt` (#28068)
danielhanchen Sep 6, 2026
465e49b
convert : add `--fuse-qkv` flag to fuse Q/K/V into QKV during HF-to-G…
JoursBleu Sep 6, 2026
8fe90e1
vulkan: add TQ1_0 support (mm, mat-vec, mat-vec-id, dequant, get_rows…
Anjielon Sep 7, 2026
2092353
ci : add container image checking and tagging (wip) (#28394)
danbev Sep 7, 2026
9ac8c40
vulkan: rms_norm fusion opportunities (#28024)
jeffbolznv Sep 7, 2026
992cb50
ggml: allow backend inputs to not create another split (#28387)
am17an Sep 7, 2026
b74f590
ggml-cuda: fix divergent barrier in f16 flash attention (#27870)
siavashnorouzi Sep 7, 2026
9a75705
convert : write explicit recurrent_layers for Qwen3-Next / Qwen3.5 (#…
devvexus Sep 7, 2026
5202104
caps : recheck typed content if template checks for string (#28511)
CISC Sep 7, 2026
1173700
examples : print ggml_version and ggml_commit in test-cmake [no ci] (…
danbev Sep 7, 2026
0cae430
vulkan: support type-aligned GET_ROWS (#28253)
jeffbolznv Sep 7, 2026
73ab759
CUDA: branchless Q4_K/Q5_K unpack to speed up mmvq, L2 prefetch on DG…
praneshgo Sep 7, 2026
d23c47f
convert : refactor Hy4-preview conversion - move HC tensor mapping to…
fairydreaming Sep 7, 2026
4735997
ggml: add gfx90c HIP support (#26454)
AuroraRAS Sep 7, 2026
0c96345
CUDA: size routed MoE MMQ N-tiles from typical expert width on RDNA3 …
ravel7524 Sep 7, 2026
7a333e7
vulkan: add DeepSeek-V4 hyper-connection fused ops (DSV4_HC_COMB/PRE/…
kh0pper Sep 7, 2026
dbeb375
sycl: add a batched L2_NORM kernel (#28222)
Titaniumtown Sep 7, 2026
160bd03
server: fix LRU hang on multiple requests same model (#28539)
ngxson Sep 7, 2026
c0b1871
webgpu: format the GET_ROWS case block (#28542)
ServeurpersoCom Sep 7, 2026
ccc3646
nix : update deprecated expressions (#28145)
ZhaolunYin Sep 7, 2026
e71b805
Revert "CUDA: size routed MoE MMQ N-tiles from typical expert width o…
pwilkin Sep 7, 2026
f114f91
tests : initialize the L2_NORM batch array (#28553)
ServeurpersoCom Sep 7, 2026
67672dc
ci : bump ty to 0.0.78 (#28548)
CISC Sep 7, 2026
050dde5
hexagon: add RELU and LEAKY_RELU ops (#28585)
tboinovski1 Sep 8, 2026
9dcf84e
model : support Kimi-K3 recurrent-state rollback (#28466)
soulmachine Sep 8, 2026
5a6caa0
ggml : update ggml_prec specification (#26675)
ggerganov Sep 8, 2026
7d701b5
opencl: properly handle non-contiguous inputs to conv2d (#28503)
lhez Sep 8, 2026
895c045
chat : split specialized parsers into common/parsers (#27764)
pwilkin Sep 8, 2026
f014bfe
Fix Vulkan-Hpp handle usage on 32-bit targets. (#22892)
miyanyan Sep 8, 2026
64e9bce
vulkan : fuse UNARY(GELU|SIGMOID|SILU|SOFTPLUS) + MUL (#27220)
Ankk98 Sep 8, 2026
ca86fb2
llama : add missing headers (#28566)
peppergrayxyz Sep 8, 2026
1744c6b
ci : add PYTEST_WORKERS=1 to fix server-self-hosted job (#28603)
danbev Sep 8, 2026
03fa73c
ci : disable npm gha cache (#28600)
CISC Sep 8, 2026
415e909
spec: single device drafter should create meta backend wrapper (#28390)
am17an Sep 8, 2026
88ada91
metal : fix idle threads in mul_mv_iq3_xxs for ne00 < 1024 (#28086)
masterFoad Sep 8, 2026
5d806aa
server : apply checkpoint min-step eviction only when the checkpoint …
masterFoad Sep 8, 2026
d4389a4
Revert "ggml-cuda : restore prop.integrated on HIP builds (#24233)" (…
IMbackK Sep 8, 2026
9113cc1
ggml : fix msvc+clang ggml_vld1q_u32 (#28284)
sarahwu185 Sep 8, 2026
f3f1a8f
llama: disable lazy tensor loading by default on iGPUs (#28326)
0cc4m Sep 8, 2026
304665f
Add IQ type handling for MoE (#28476)
cwriter Sep 9, 2026
30b6a75
tests : use less threads for data initialization (#28325)
ggerganov Sep 9, 2026
1945e09
vulkan: add f16 B-type matmul pipelines and warp tile size tuning for…
fish-jiang Sep 9, 2026
df750f7
vulkan: add dedicated iq4_xs mat-vec shader (#28426)
WakeUpMorty Sep 9, 2026
b31b71f
jinja: treat a null left operand of in as a plain lookup (#28620)
ServeurpersoCom Sep 9, 2026
6de9cdb
mtmd: propagate video ID to bitmap (#28601)
ngxson Sep 9, 2026
e2d2c0d
model: fix granite3 moe unknown parameter count (#28632)
taronaeo Sep 9, 2026
14a9d09
args: officially deprecate --mmap|mlock|dio (#28334)
taronaeo Sep 9, 2026
5a4d0fe
CUDA: replace GGML_FA_ALL_QUANTS with GGML_FA_QUANTS, more control ov…
pwilkin Sep 9, 2026
d4abd57
CUDA: size routed MoE MMQ N-tiles from typical expert width on RDNA3 …
pwilkin Sep 9, 2026
4850c77
llama : use int32_t for llama_sampler_chain_n return type (#28631)
linsen458-spec Sep 9, 2026
9cf3bf2
py : bump numpy to 2.4.6 (#28649)
CISC Sep 9, 2026
4b98ab8
py : lower numpy to 2.2.6 (#28654)
CISC Sep 9, 2026
22397c3
vulkan: Convert FILL to distribute workgroups in 2D to avoid exceedin…
rillomas Sep 9, 2026
6d9c82e
hexagon: rope updates (#28628)
tboinovski1 Sep 9, 2026
91f6a6c
vulkan: use spec constant for matrix matrix multiplication A-type (#2…
0cc4m Sep 9, 2026
434ddbb
ci: fix sanitizer tests (#28583)
netrunnereve Sep 9, 2026
d7e8643
model: fix all granite family parameter counts (#28643)
taronaeo Sep 10, 2026
f1b6fbf
ggml-cpu(s390x): add Q1_0 vector intrinsic support (#28606)
taronaeo Sep 10, 2026
4ea6d1b
ggml-cpu(s390x): add repack support for q4_0 (#28667)
taronaeo Sep 10, 2026
72797e8
vulkan : add command-buffer debug labels for GPU profilers (#28101)
CordellBlanchard Sep 10, 2026
311d421
memory : avoid allocating V cache for indexer (it's not used) (#28330)
fairydreaming Sep 10, 2026
8c322d5
convert : expand Nemotron H conversion fix (#28689)
frozenblade1224 Sep 10, 2026
3ff67eb
vulkan: fall back to shared-memory reduction for dmmv on PowerVR (#28…
JulianPscheid Sep 10, 2026
e5a8d43
tests : drop SYCL special-casing in test-backend-ops.cpp (#28688)
ggerganov Sep 10, 2026
c32d1da
tests : increase tolerance for Add fusion tests (#28691)
gaugarg-nv Sep 10, 2026
d344123
models: clean up some dead switch branches in old models (#28669)
iggy Sep 10, 2026
41fc758
scripts : use sed instead of grep for version parsing [no ci] (#28700)
danbev Sep 10, 2026
fa67698
spec: fix failed to decode mtmd chunk with DFlash (#28587)
jesdga95 Sep 10, 2026
18c17b4
ci : Update WoA CUDA 13.4 release to use 13.4.1 GA redistributables (…
shivamkumard-ctrl Sep 10, 2026
52d4268
ci : add self-hosted-gpu-cuda and server-sanitize to hf-jobs (#28693)
CISC Sep 10, 2026
6788edb
vulkan: small M matrix optimizations for qwen (#28457)
jeffbolznv Sep 10, 2026
50182a5
vulkan: use add_alloc_dep to enable topk_moe fusion for prefill (#28422)
jeffbolznv Sep 10, 2026
28ff095
vulkan: use CPU writes in ggml_backend_vk_cpy_tensor_async if the con…
jeffbolznv Sep 10, 2026
df03399
opencl: add A8 Q4_0 mm binary kernel support (#28268)
shaofeiqi Sep 10, 2026
481c65f
vulkan: fix data race and OOB access in argsort(large) (#28705)
jeffbolznv Sep 11, 2026
451b89b
ci : key cache to sanitizer matrix (#28708)
CISC Sep 11, 2026
16378d9
CUDA/HIP: Flash Attention tuning (gfx1201) (#28102)
pwilkin Sep 11, 2026
b0dcb81
server: fix speculation after an image (#28715)
jesdga95 Sep 11, 2026
5cdd3d1
model : fix MTP context kv cache allocation for deepseek2, glm4moe, ……
LoganChu Sep 11, 2026
aac8102
metal : fix idle threads in the remaining iq mul_mv kernels for ne00 …
masterFoad Sep 11, 2026
a2878d3
metal : single-source fusion table + fusion debug rework (#28164)
ggerganov Sep 11, 2026
1dfe94e
common : fix typo in speculative.cpp comment [no ci] (#28750)
danbev Sep 11, 2026
3bcfeb7
cmake : add PCH and unity build to improve build times (#28091)
danbev Sep 11, 2026
5bda51b
metal : skip the empty half of the mul_mm_id token tile (#28301)
masterFoad Sep 11, 2026
43f3dda
ggml: skip 0-sized ids tensor when offloading selected experts (#28739)
am17an Sep 11, 2026
8172e65
tests: tolerate a shared pool abort in test_completion_unified (#28759)
ServeurpersoCom Sep 11, 2026
982937a
tests: extend test-quantize-fns to test nrc=2 (i8mm) kernels (#16234)
Rohanjames1997 Sep 11, 2026
b78a39a
ci : run test-backend-ops as a dedicated ci/run.sh test (#28740)
ggerganov Sep 11, 2026
8ea2902
cmake : skip PCH for llama-server PCH when using MSVC (#28763)
danbev Sep 11, 2026
82d6bb2
server: refactor subproc handling (#28555)
ngxson Sep 11, 2026
d3146f2
ggml-webgpu: Update to a recent version of Dawn (#28683)
MendyBerger Sep 12, 2026
eafe15a
hexagon: support for multi-device model split (aka row-split) (#28589)
max-krasnyansky Sep 12, 2026
3f5e94d
webgpu: align tensor bindings to the type block size (#28382)
ServeurpersoCom Sep 12, 2026
07fc977
opencl: add bin kernel `kernel_gemm_noshuffle_q4_k_f32_32b_trans_ila_…
shaofeiqi Sep 12, 2026
8a56aed
opencl: fix several bugs where the backend aborts (#27630)
wanghqc Sep 12, 2026
c069aa7
server: frame the router child state command as a whole line (#28747)
ServeurpersoCom Sep 12, 2026
f3a33df
rpc : fix linking when compiling with BUILD_SHARED_LIBS=OFF (#28492)
EAddario Sep 12, 2026
2a3005c
syscl : Handle (fail gracefully) unsupported tq1_0 quants (#28681)
mctylr-gh Sep 12, 2026
718f7b4
vendor : update cpp-httplib to 0.56.0 (#28787)
cabelo Sep 12, 2026
e192abb
server : add missing headers (#28795)
angt Sep 12, 2026
c8edceb
ggml-cuda: hip add specific config table for AMD GCN (#27841)
thelittlefireman Sep 12, 2026
56381e4
server : allow model downloads at model limit fix issue #26809 (#28530)
chenmiaoming Sep 12, 2026
3057bb6
ui : add cache (#28802)
angt Sep 12, 2026
737e098
cmake: leave the timestamp out of precompiled headers on clang (#28816)
ServeurpersoCom Sep 12, 2026
ae9afff
jinja : support dot property integer literals (#28817)
CISC Sep 12, 2026
acecd56
common : implement common_schema internal representation for JSON sch…
aldehir Sep 12, 2026
8e33095
common: add LOG_JSON macro to log structured data (#28586)
ngxson Sep 12, 2026
790cf51
chat : improve parsing of complex types in qwen3-coder (#28742)
aldehir Sep 13, 2026
56b9eb2
opencl: apply the noshuffle row-alignment rule to q4_K, q5_K and q8_0…
wanghqc Sep 13, 2026
f1e44dc
vulkan: workaround NV queuesubmit driver bug (#28830)
jeffbolznv Sep 13, 2026
002a12a
ci : cap test-backend-ops parallel jobs at 2 and add a 3600s timeout …
ggerganov Sep 13, 2026
37b3a9e
ci : remove leftover command (#28839)
CISC Sep 13, 2026
4a89937
tests : reduce FA test sizes (#28842)
ggerganov Sep 13, 2026
bc52a12
pi : prefer PI_MODEL_NAME env var for model disclosure (#28853)
ggerganov Sep 13, 2026
c95f8e4
ci : run editorconfig and code-style checks on ubuntu-slim (#28854)
ggerganov Sep 13, 2026
b6b003d
sycl : Fix get mem error (#28227)
arthw Sep 13, 2026
243a308
tests : fix typo in test-quant-type-selection for nemotron 3 nano (#2…
mctylr-gh Sep 13, 2026
6978052
ggml-cpu(s390x): guard VXE-only repack helpers (#28775)
bernardladenthin Sep 13, 2026
e49d2c2
models : guard the expert FFN size fallback in nemotron-h against a z…
YanissAmz Sep 13, 2026
5f436dd
tests : exclude HY_V4 from WebGPU test-llama-archs tests (#28855)
fairydreaming Sep 13, 2026
7a16a6c
grammar : coalesce find + insert into a single insert and adjust move…
HanClinto Sep 13, 2026
ad6c668
ggml-cuda: fallback to F32 on device without BF16 hardware accelerati…
thelittlefireman Sep 13, 2026
093a2f8
common : move llama_n_rs_seq to before llama_decode (#28749)
danbev Sep 14, 2026
661643e
sycl : fix oneDNN scratchpad breaking the pool free order (#28704)
lslusarczyk Sep 14, 2026
15d8f2d
ci : remove gg_sum summary logic (#28857)
ggerganov Sep 14, 2026
89fe242
ci : trigger self-hosted CI on changes to ci/run.sh (#28859)
ggerganov Sep 14, 2026
2f53959
ggml-cpu : disable PCH and fix CACHE_LINE_SIZE ambiguity to fix heap …
ggerganov Sep 14, 2026
21f6b0d
sycl: rfc: Use radix select for top_k (#28670)
cwriter Sep 14, 2026
3d10bcd
llama: add Maple 20B-A1B ternary MoE architecture (CPU) (#27000)
AlexGabbia Sep 14, 2026
be2c6d7
tests(s390x): add non-vxe build to tests (#28776)
taronaeo Sep 14, 2026
1aca1f9
models : fix mimo2 swa pattern load (#28865)
CISC Sep 14, 2026
97e4ca7
models : fix incorrect uses of get_key_or_arr (#28868)
CISC Sep 14, 2026
bbdd9f2
tests : add fusion baseline README and broaden fusion CI triggers (#2…
ggerganov Sep 14, 2026
eeea731
ggml : bump version to 0.24.0 (ggml/1627)
ggerganov Sep 14, 2026
d9e03f1
sync : ggml
ggerganov Sep 14, 2026
b29c606
llama.cpp : bump version to 0.4.1 (#28900)
ggerganov Sep 14, 2026
dfe4516
scripts: Add script to verify API/ABI compatibility (#28579)
ckastner Sep 14, 2026
f3a184b
cmake : remove precompiled headers (#28892)
danbev Sep 14, 2026
b4fa47d
release : added gfx1103 to ubuntu rocm build (#28423)
apparle Sep 14, 2026
41abbfd
qwen4exp: enable rms_norm + mul fusion (#28896)
am17an Sep 14, 2026
391fac1
ci : add ubuntu-cuda builds to release (#28186)
ORippler Sep 14, 2026
bfdc321
HIP: fattn-mma: use fp32 accumulation on MFMA devices (#28576)
IMbackK Sep 14, 2026
96ffdc4
CI: hip-quality-check: ignore spill added in bfdc32183d57f1e35bacf35c…
IMbackK Sep 14, 2026
7cf1c54
ci : reuse build tag name when used instead of safe one (#28911)
CISC Sep 14, 2026
1bc7a5a
webui: stop re-probing disabled /tools endpoint on every message (#28…
geckguy Sep 14, 2026
69eb250
cmake : use PROJECT_SOURCE_DIR instead of CMAKE_SOURCE_DIR (#28771)
danbev Sep 15, 2026
4c9233c
cuda : enable i16 and i32 for DUP (#28897)
amankarki151 Sep 15, 2026
987498f
ci : fix android release (#28936)
CISC Sep 15, 2026
0ecb159
ci: Bump CUDA Windows x64 builds to 13.4.1 (#28930)
shivamkumard-ctrl Sep 15, 2026
1e7bcf3
metal : add FA kernels for HSK=96, HSV=64 (MiniCPM3) (#28599)
wyanzhao Sep 15, 2026
1af6c65
ci: bump kleidiai runners from 22.04 to 24.04 (#28885)
taronaeo Sep 15, 2026
6ec1a7e
opencl: add generic ssm_scan (#28881)
lhez Sep 15, 2026
77d554b
OpenVINO: optimize stateful decode and GPU MoE inference (#28638)
wine99 Sep 15, 2026
fc82583
vulkan: support sparse Flash Attention (#28105)
0cc4m Sep 15, 2026
9e71716
models : move build_arch_graph() after graph() template specializatio…
cpeterso Sep 15, 2026
5431581
cuda: support row-contiguous SUM_ROWS (#26308)
MohamedElashri Sep 15, 2026
7609846
rpc : hash-cache only weights (#28789)
Patt92 Sep 15, 2026
6011c34
docs: Rule of thumb for AI review time [no ci] (#28945)
JohannesGaessler Sep 15, 2026
d1d3c33
ci: build MUSA for only 1 arch (#28944)
am17an Sep 15, 2026
9f31776
opencl: choose the MoE expert matmul by batch size for speculative de…
wanghqc Sep 15, 2026
38a5b42
HIP: Enable AllReduce for ROCm (#27825)
Stastez Sep 15, 2026
72b590d
hex-cpy: use dma if src and dst are contiguous (#28906)
trivikram-reddy1 Sep 15, 2026
930e2fa
hexagon: add back missing contiguous fast-path and hvx_copy_uu for ea…
jhen0409 Sep 15, 2026
e13469a
llama-bench: support --version to print build info (#28971)
asbelin Sep 16, 2026
583926e
ci : add self-hosted webgpu to hf-jobs (#28712)
CISC Sep 16, 2026
0a8b29a
metal: fix NaN in mul_mm_id when activations exceed f16 range (#26223)
mdegans Sep 16, 2026
d4365d9
vulkan: make MUL_MAT_ID BN/2 tail unconditional (#28923)
SG-Amadeus Sep 16, 2026
0bec16e
chat : force `\n</think>` on reasoning budget end for qwen3-coder (#2…
aldehir Sep 16, 2026
fccf716
HIP: broaden MoE ncols_opt tile heuristic on RDNA3.5 architecture (#2…
WenqiangJia2026 Sep 16, 2026
37b53fd
qwen4exp: add hc ops (#28901)
am17an Sep 16, 2026
b04d4e5
Change max context length for auto-fitting with unified KV (#28849)
gaugarg-nv Sep 16, 2026
6019933
rpc : invalidate cached compute graph when a referenced buffer is fre…
y198nt Sep 16, 2026
f266648
spacemit : fix wrong transpose function for int16 data (#25161)
I3eg1nner Sep 16, 2026
83078fe
CUDA/HIP: improve access patterns in im2col (#28013)
IMbackK Sep 16, 2026
7d6f5d0
model : add support for HrmTextForCausalLM (DFM Mimir 1B) (#27625)
noctrex Sep 16, 2026
7ceed87
models : allow Nemotron-H models to only define layer_norm_epsilon (#…
frozenblade1224 Sep 16, 2026
82324fc
hexagon: accept the zeroed rope probe in supports_op (#28995)
rjtokenring Sep 16, 2026
1ec8188
hexagon: Support for K-Quants Q4_K and Q6_K (#28994)
rjtokenring Sep 16, 2026
2f3fd02
Enable CUDA graph for MTP draft (#28549)
gaugarg-nv Sep 16, 2026
c6824a9
ci: switch fast jobs back to github (#28959)
netrunnereve Sep 16, 2026
fb27a52
TP: fix split state and granularity for fused QKV gemma4, qwen35 (#28…
dfriehs Sep 16, 2026
4bc272f
vulkan: work around NV bug with argsort_large.comp (#28975)
jeffbolznv Sep 16, 2026
aa39d7a
[SYCL] Fix function signature for `ggml_backend_sycl_split_buffer_typ…
mctylr-gh Sep 17, 2026
35822af
vulkan: support qwen4exp hc ops (#28988)
0cc4m Sep 17, 2026
05f2dcf
vulkan: fix buffer_reference alignment in im2col shaders (#28996)
apollo-2006 Sep 17, 2026
79bfc1d
docs: remove JG as CODEOWNER for test-llama-archs (#29003)
JohannesGaessler Sep 17, 2026
c57da6f
opencl: fix various warnings (#28984)
lhez Sep 17, 2026
817e5f8
sycl: ssm_conv: fuse the SiLU epilogue into the ssm_conv kernel (#28929)
Titaniumtown Sep 17, 2026
7490357
vulkan: skip unneeded MoE work in mul_mm coopmat1 path (#25483)
fish-jiang Sep 17, 2026
c9a5eee
sycl : fix the B70 mem allocate error when >19.3GB (#28953)
arthw Sep 17, 2026
81aeaeb
gguf : align the data section relative to the GGUF start, not the fil…
ykhrustalev Sep 17, 2026
7f6f0c2
chat : add message delimiters to the DeepSeek V3.2/V4 parser (#29008)
midagedev Sep 17, 2026
87f9c82
ci : add API/ABI check to make-release workflow [no ci] (#28947)
danbev Sep 17, 2026
f172be7
vulkan: split buffers and debug code into separate files, add shared …
0cc4m Sep 17, 2026
4ff829e
ui: fix removed reasoning menu in single model mode on desktop (#27985)
benbenbenbenbenben Sep 17, 2026
ebbb185
openvino : Update OpenVINO to 2026.4;fix clangd,MSVC warnings; (#29009)
ravi9 Sep 17, 2026
7076180
model : extend Nemotron MTP support (#29018)
frozenblade1224 Sep 17, 2026
b49650a
model : skip gate_up_exps if TENSOR_SKIP is set (#29014)
dfriehs Sep 17, 2026
c77ae69
rpc : skip ACCEL devices (#29020)
pcuenca Sep 17, 2026
972d231
ci : add missing evict-old-files (#29041)
CISC Sep 17, 2026
6ea0bbe
Merge upstream/master into fork master
Sep 19, 2026
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
The table of contents is too big for display.
Diff view
Diff view
  •  
  •  
  •  
12 changes: 6 additions & 6 deletions .devops/nix/package.nix
Original file line number Diff line number Diff line change
Expand Up @@ -31,7 +31,7 @@
]
&& blas.meta.available,
useCuda ? config.cudaSupport,
useMetalKit ? stdenv.isAarch64 && stdenv.isDarwin,
useMetalKit ? stdenv.hostPlatform.isAarch64 && stdenv.hostPlatform.isDarwin,
# Increases the runtime closure size by ~700M
useMpi ? false,
useRocm ? config.rocmSupport,
Expand Down Expand Up @@ -92,7 +92,7 @@ let

cudaBuildInputs = with cudaPackages; [
cuda_cudart
cuda_cccl # <nv/target>
cccl # <nv/target>
libcublas
];

Expand Down Expand Up @@ -166,7 +166,7 @@ effectiveStdenv.mkDerivation (finalAttrs: {
# `xcrun` is used find the path of the Metal compiler, which is varible
# and not on $PATH
# see https://github.com/ggml-org/llama.cpp/pull/6118 for discussion
__noChroot = effectiveStdenv.isDarwin && useMetalKit && precompileMetalShaders;
__noChroot = effectiveStdenv.hostPlatform.isDarwin && useMetalKit && precompileMetalShaders;

nativeBuildInputs =
[
Expand All @@ -181,10 +181,10 @@ effectiveStdenv.mkDerivation (finalAttrs: {
autoAddDriverRunpath
]
++ optionals (effectiveStdenv.hostPlatform.isGnu && enableStatic) [ glibc.static ]
++ optionals (effectiveStdenv.isDarwin && useMetalKit && precompileMetalShaders) [ xcrunHost ];
++ optionals (effectiveStdenv.hostPlatform.isDarwin && useMetalKit && precompileMetalShaders) [ xcrunHost ];

buildInputs =
optionals effectiveStdenv.isDarwin darwinBuildInputs
optionals effectiveStdenv.hostPlatform.isDarwin darwinBuildInputs
++ optionals useCuda cudaBuildInputs
++ optionals useMpi [ mpi ]
++ optionals useRocm rocmBuildInputs
Expand Down Expand Up @@ -245,7 +245,7 @@ effectiveStdenv.mkDerivation (finalAttrs: {

# Configurations that are known to result in build failures. Can be
# overridden by importing Nixpkgs with `allowBroken = true`.
broken = (useMetalKit && !effectiveStdenv.isDarwin);
broken = (useMetalKit && !effectiveStdenv.hostPlatform.isDarwin);

description = "Inference of LLaMA model in pure C/C++${descriptionSuffix}";
homepage = "https://github.com/ggml-org/llama.cpp/";
Expand Down
23 changes: 13 additions & 10 deletions .devops/openvino.Dockerfile
Original file line number Diff line number Diff line change
@@ -1,18 +1,18 @@
ARG OPENVINO_VERSION_MAJOR=2026.2.1
ARG OPENVINO_VERSION_FULL=2026.2.1.21919.ede283a88e3
ARG OPENVINO_VERSION_MAJOR=2026.4
ARG OPENVINO_VERSION_FULL=2026.4.0.22959.99c81491cc3
ARG UBUNTU_VERSION=24.04

# Intel GPU driver versions. https://github.com/intel/compute-runtime/releases
ARG IGC_VERSION=v2.36.3
ARG IGC_VERSION_FULL=2_2.36.3+21719
ARG COMPUTE_RUNTIME_VERSION=26.22.38646.4
ARG COMPUTE_RUNTIME_VERSION_FULL=26.22.38646.4-0
ARG IGC_VERSION=v2.40.13
ARG IGC_VERSION_FULL=2_2.40.13+22418
ARG COMPUTE_RUNTIME_VERSION=26.31.39395.13
ARG COMPUTE_RUNTIME_VERSION_FULL=26.31.39395.13-0
ARG IGDGMM_VERSION=22.10.0

# Intel NPU driver versions. https://github.com/intel/linux-npu-driver/releases
ARG NPU_DRIVER_VERSION=v1.33.0
ARG NPU_DRIVER_FULL=v1.33.0.20260529-26625960453
ARG LIBZE1_VERSION=1.27.0-1~24.04~ppa2
ARG NPU_DRIVER_VERSION=v1.38.0
ARG NPU_DRIVER_FULL=v1.38.0.20260910-34487311128
ARG LIBZE1_VERSION=1.32.0-1~24.04~ppa1

# Optional proxy build arguments
ARG http_proxy=
Expand Down Expand Up @@ -90,6 +90,9 @@ RUN bash -c "source ${OpenVINO_DIR}/setupvars.sh && \
cmake -B build/ReleaseOV -G Ninja \
-DCMAKE_BUILD_TYPE=Release \
-DLLAMA_BUILD_TESTS=OFF \
-DGGML_NATIVE=OFF \
-DGGML_BACKEND_DL=ON \
-DGGML_CPU_ALL_VARIANTS=ON \
-DGGML_OPENVINO=ON && \
cmake --build build/ReleaseOV --parallel "

Expand Down Expand Up @@ -170,7 +173,7 @@ RUN --mount=type=cache,target=/var/cache/intel-npu,sharing=locked \
fi; \
DEB=/var/cache/intel-npu/libze1_${LIBZE1_VERSION}_amd64.deb; \
if [ ! -f "$DEB" ]; then \
wget -q -O "$DEB" https://snapshot.ppa.launchpadcontent.net/kobuk-team/intel-graphics/ubuntu/20260324T100000Z/pool/main/l/level-zero-loader/libze1_${LIBZE1_VERSION}_amd64.deb; \
wget -q -O "$DEB" https://snapshot.ppa.launchpadcontent.net/kobuk-team/intel-graphics/ubuntu/20260830T100000Z/pool/main/l/level-zero-loader/libze1_${LIBZE1_VERSION}_amd64.deb; \
fi; \
mkdir /tmp/npu/ && cd /tmp/npu/ && tar -xf "$TGZ" && cp "$DEB" .; \
apt-get update; \
Expand Down
2 changes: 1 addition & 1 deletion .ecrc
Original file line number Diff line number Diff line change
@@ -1,5 +1,5 @@
{
"Exclude": ["^\\.gitmodules$", "stb_image\\.h"],
"Exclude": ["^\\.gitmodules$", "stb_image\\.h", "examples/test-cmake/build/", "examples/test-cmake/build-subdir/"],
"Disable": {
"IndentSize": true
}
Expand Down
2 changes: 1 addition & 1 deletion .github/ISSUE_TEMPLATE/config.yml
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
blank_issues_enabled: true
blank_issues_enabled: false
contact_links:
- name: Got an idea?
url: https://github.com/ggml-org/llama.cpp/discussions/categories/ideas
Expand Down
95 changes: 95 additions & 0 deletions .github/actions/ccache-buckets/action.yml
Original file line number Diff line number Diff line change
@@ -0,0 +1,95 @@
name: "ccache-buckets"
description: "Save/restore latest GitHub Actions ccache matching a key prefix to/from HF buckets"
inputs:
key:
description: "Cache key prefix to match and load"
required: true
folder:
description: "Bucket folder containing ccache files"
required: true
evict-old-files:
description: "Corresponds to the ccache --evict-older-than AGE option, where AGE is the number of seconds or days followed by the 's' or 'd' suffix respectively."
default: ''
save:
description: "Save ccache"
required: false
default: false
type: boolean
hf_bucket:
description: 'Hugging Face buckets path'
required: true

runs:
using: "composite"
steps:
- name: Install Hugging Face Hub CLI
shell: bash
run: |
python3 -m venv .venv-hf
.venv-hf/bin/pip install -U huggingface_hub==1.28.0

- name: Restore ccache from buckets
if: ${{ inputs.save != 'true' }}
shell: bash
run: |
set +e -uo pipefail
source .venv-hf/bin/activate
CCACHE_DIR=$(ccache -k cache_dir)
if [[ -d "$CCACHE_DIR" ]]; then
CACHE_PATH=$(hf buckets list "hf://buckets/${{ inputs.hf_bucket }}/${{ inputs.folder }}" --json | jq -r '[.[] | select(.type == "file") | select(.path | startswith("${{ inputs.folder }}/${{ inputs.key }}") and endswith(".tar.gz"))] | sort_by(.path) | last | .path // ""')
if [[ -n "$CACHE_PATH" ]]; then
echo "Restoring ccache from '$CACHE_PATH'."
hf buckets cp "hf://buckets/${{ inputs.hf_bucket }}/$CACHE_PATH" ccache_bucket.tar.gz
mkdir -p ccache_bucket
if tar -xzf ccache_bucket.tar.gz -C ccache_bucket; then
rm -rf "$CCACHE_DIR"
mv ccache_bucket "$CCACHE_DIR"
ccache -z
fi
rm ccache_bucket.tar.gz
else
echo "No ccache found."
fi
else
echo "'$CCACHE_DIR' not found."
fi

- name: Save ccache to buckets
if: ${{ inputs.save == 'true' }}
shell: bash
run: |
if [[ -n "$HF_TOKEN" ]]; then
set +e -uo pipefail
source .venv-hf/bin/activate
CCACHE_DIR=$(ccache -k cache_dir)
if [[ -d "$CCACHE_DIR" ]]; then
ccache -s
if [[ -n "${{ inputs.evict-old-files }}" ]]; then
ccache --evict-older-than "${{ inputs.evict-old-files }}"
fi
DATESTAMP=$(date -u +'%Y-%m-%dT%H:%M:%SZ')
CACHEFILE="${{ inputs.key }}-$DATESTAMP.tar.gz"
if tar -czf ccache_bucket.tar.gz -C "$CCACHE_DIR" .; then
hf buckets cp ccache_bucket.tar.gz "hf://buckets/${{ inputs.hf_bucket }}/${{ inputs.folder }}/$CACHEFILE"
fi
rm ccache_bucket.tar.gz
else
echo "'$CCACHE_DIR' not found."
fi
fi

- name: Remove old ccache files from buckets
if: ${{ inputs.save == 'true' }}
shell: bash
run: |
if [[ -n "$HF_TOKEN" ]]; then
set +e -uo pipefail
source .venv-hf/bin/activate
CACHE_FILES=$(hf buckets list "hf://buckets/${{ inputs.hf_bucket }}/${{ inputs.folder }}" --json | jq -r '[.[] | select(.type == "file") | select((.uploaded_at | .[:19]+"Z" | fromdateiso8601) < (now - 5 * 60)) | select(.path | startswith("${{ inputs.folder }}/${{ inputs.key }}") and endswith(".tar.gz"))] | sort_by(.path)[:-1] | .[] | [.path // ""] | @tsv')
if [[ -n "$CACHE_FILES" ]]; then
echo "Removing old ccache files..."
while IFS=$'\t' read -r CACHE_PATH; do
hf buckets rm "hf://buckets/${{ inputs.hf_bucket }}/$CACHE_PATH" -y
done <<< "$CACHE_FILES"
fi
fi
48 changes: 38 additions & 10 deletions .github/actions/ccache-clear/action.yml
Original file line number Diff line number Diff line change
@@ -1,22 +1,50 @@
# note: place this as the last step of the job, so the new cache is saved by "Post ccache" right after the old one is cleared
name: "ccache-clear"
description: "Delete all GitHub Actions caches matching a key prefix"
description: "Delete GitHub Actions caches matching a key prefix, oldest first"
inputs:
key:
description: "Cache key prefix to match and delete"
required: true
older:
description: "Only delete caches created more than this long ago (e.g. 90m, 1h, 1d). By default all matching caches are deleted"
required: false
default: ""
min:
description: "Stop deleting if fewer than this many caches would remain (e.g. 1). By default there is no minimum"
required: false
default: "0"
dry-run:
description: "Only print the caches that would be deleted, without deleting them"
required: false
default: "false"

runs:
using: "composite"
steps:
- name: Clear caches
- name: Install GitHub CLI if missing
shell: bash
run: |
CACHES=$(gh cache list --key "ccache-${{ inputs.key }}" --json id,key --jq '.[] | "\(.id) \(.key)"' 2>/dev/null)
if [ -z "$CACHES" ]; then
echo "No caches found with key prefix: ${{ inputs.key }}"
exit 0
# e.g. in container jobs, where it is not preinstalled
if ! command -v gh >/dev/null 2>&1; then
echo "GitHub CLI not found, installing..."
if ! command -v curl >/dev/null 2>&1; then
apt-get update >/dev/null 2>&1 || true
apt-get install -y curl >/dev/null 2>&1 || true
fi
mkdir -p -m 755 /etc/apt/keyrings
curl -fsSL https://cli.github.com/packages/githubcli-archive-keyring.gpg | tee /etc/apt/keyrings/githubcli-archive-keyring.gpg >/dev/null
chmod go+r /etc/apt/keyrings/githubcli-archive-keyring.gpg
echo "deb [arch=$(dpkg --print-architecture) signed-by=/etc/apt/keyrings/githubcli-archive-keyring.gpg] https://cli.github.com/packages stable main" > /etc/apt/sources.list.d/github-cli.list
apt-get update >/dev/null 2>&1 || true
apt-get install -y gh || { echo "Failed to install GitHub CLI (gh)" >&2; exit 1; }
fi
while read -r id key; do
echo "Deleting cache: $id ($key)"
gh cache delete "$id"
done <<< "$CACHES"
command -v gh >/dev/null 2>&1 || { echo "GitHub CLI (gh) is required but could not be installed" >&2; exit 1; }

- name: Clear caches
shell: bash
run: |
bash scripts/ccache-clear.sh \
--key "${{ inputs.key }}" \
--older "${{ inputs.older }}" \
--min "${{ inputs.min }}" \
${{ inputs.dry-run == 'true' && '--dry-run' || '' }}
2 changes: 1 addition & 1 deletion .github/actions/get-tag-name/action.yml
Original file line number Diff line number Diff line change
Expand Up @@ -14,7 +14,7 @@ runs:
run: |
BUILD_NUMBER="$(git rev-list --count HEAD)"
SHORT_HASH="$(git rev-parse --short=7 HEAD)"
if [[ "${{ env.BRANCH_NAME }}" == "master" ]]; then
if [[ "${{ env.BRANCH_NAME }}" == "master" || "${{ env.BRANCH_NAME }}" == "b${BUILD_NUMBER}" ]]; then
echo "name=b${BUILD_NUMBER}" >> $GITHUB_OUTPUT
else
SAFE_NAME=$(echo "${{ env.BRANCH_NAME }}" | tr '/' '-')
Expand Down
20 changes: 0 additions & 20 deletions .github/actions/linux-setup-vulkan/action.yml

This file was deleted.

Loading
Loading