Commits

Commit history

1. Commits on Sep 30, 2026

   • musa : define __CUDA_ARCH__ for device passes (#29508)

     Show description for 272aad8
     yeahdongcn

     authored
     272aad8View commit details

     Copy full SHA for 272aad8

     Browse repository at this point

   • ci : add models backend check (#29651)

     Show description for 72db1e0
     CISC

     authored
     72db1e0View commit details

     Copy full SHA for 72db1e0

     Browse repository at this point

   • SYCL: reduce tensor allreduce sync with pinned host buffers (#29604)

     Captain-Tripps

     authored
     2a53aceView commit details

     Copy full SHA for 2a53ace

     Browse repository at this point

   • add GLM-5.3-Flash (GLM5-Next) support (#27773)

     Show description for 649dcb1
     

     4 peopleauthored
     649dcb1View commit details

     Copy full SHA for 649dcb1

     Browse repository at this point

   • vendor: update BoringSSL to 0.20260929.0 (#29669)

     cabelo

     authored
     931351eView commit details

     Copy full SHA for 931351e

     Browse repository at this point

   • ggml-zdnn: impl buffer reset, fix memory leaks (#29637)

     Show description for eae11d2
     taronaeo

     authored
     eae11d2View commit details

     Copy full SHA for eae11d2

     Browse repository at this point
2. Commits on Sep 29, 2026

   • Hexagon f16 activation ops (#29209)

     Show description for 19e28a2
     

     cqderek

     and

     max-krasnyansky

     authored
     19e28a2View commit details

     Copy full SHA for 19e28a2

     Browse repository at this point

   • gguf : reject tensor size that wraps after padding (#26979)

     Show description for a6ea155
     x14ngch3n

     authored
     a6ea155View commit details

     Copy full SHA for a6ea155

     Browse repository at this point

   • ggml : check row bounds in get_rows_back (#29575)

     Show description for d3954b9
     angt

     authored
     d3954b9View commit details

     Copy full SHA for d3954b9

     Browse repository at this point

   • model : support classifier_pooling for rerankers (#29627)

     Show description for 48de2a1
     boshjerns

     authored
     48de2a1View commit details

     Copy full SHA for 48de2a1

     Browse repository at this point

   • hexagon: optimize concat op (#29673)

     Show description for 7fee178
     trivikram-reddy1

     authored
     7fee178View commit details

     Copy full SHA for 7fee178

     Browse repository at this point

   • CUDA: bitonic argsort handles rows wider than one block (#28957)

     Show description for 6a2743f
     ServeurpersoCom

     authored
     6a2743fView commit details

     Copy full SHA for 6a2743f

     Browse repository at this point

   • ggml-cuda: HIP: optimize packed byte subtraction (`__vsubss4` -> `__vsub4`) (#29478)

     Show description for 748d422
     

     thelittlefireman

     and

     IMbackK

     authored
     748d422View commit details

     Copy full SHA for 748d422

     Browse repository at this point

   • ci: add zdnn backend build but not test (#29541)

     Show description for cee37ff
     taronaeo

     authored
     cee37ffView commit details

     Copy full SHA for cee37ff

     Browse repository at this point

   • opencl: fix get_tensor for q5_K adreno gemm_nonshuffle kernel (#29555)

     lhez

     authored
     6dbbac4View commit details

     Copy full SHA for 6dbbac4

     Browse repository at this point

   • vulkan: Tune GDN kernel, fix Intel performance (#29476)

     Show description for 5c200e0
     0cc4m

     authored
     5c200e0View commit details

     Copy full SHA for 5c200e0

     Browse repository at this point

   • vulkan : Load F32 A matrix 2 at a time when its 2-aligned (#29254)

     Show description for 83dd71f
     TheBlueMatt

     authored
     83dd71fView commit details

     Copy full SHA for 83dd71f

     Browse repository at this point

   • vulkan: MOE aware mat_mul_id tile selection (#29182)

     Show description for 94a0ae3
     Ankk98

     authored
     94a0ae3View commit details

     Copy full SHA for 94a0ae3

     Browse repository at this point

   • ggml : fix c++ odr by properly using GGML_COMMON_DECL_CPP (#29504)

     Show description for da89bb3
     

     3 peopleauthored
     da89bb3View commit details

     Copy full SHA for da89bb3

     Browse repository at this point

   • vocab : keep </s> NORMAL in PLaMo-2 and PLaMo-3 (#29580)

     Show description for a3f84fa
     tokinasin

     authored
     a3f84faView commit details

     Copy full SHA for a3f84fa

     Browse repository at this point

   • ggml : accumulate f16 dot products in f32 on AVX512-FP16 (#29545)

     Show description for 284153e
     angt

     authored
     284153eView commit details

     Copy full SHA for 284153e

     Browse repository at this point

   • ggml : require input tensors to be GGML_OP_NONE (#29647)

     ggerganov

     authored
     b5cf8ceView commit details

     Copy full SHA for b5cf8ce

     Browse repository at this point

   • hexagon: add FP32 GELU_ERF and GEGLU_ERF support (#29631)

     Show description for e904318
     aparmp-quic

     authored
     e904318View commit details

     Copy full SHA for e904318

     Browse repository at this point

   • common : stop accepting draft tokens at EOG (#29638)

     Show description for d280808
     ethanhq

     authored
     d280808View commit details

     Copy full SHA for d280808

     Browse repository at this point

   • server : remove the built-in UI's service worker when the UI is not served (#29565)

     Show description for ba0ba54
     erusev

     authored
     ba0ba54View commit details

     Copy full SHA for ba0ba54

     Browse repository at this point

   • common : use fs::path for config dir (#29649)

     Show description for 00af635
     angt

     authored
     00af635View commit details

     Copy full SHA for 00af635

     Browse repository at this point

   • tests : adjust server string regex to also match m2 utlra results (#29648)

     ggerganov

     authored
     c85b92cView commit details

     Copy full SHA for c85b92c

     Browse repository at this point

   • common : add fs_write_atomic() (#29642)

     Show description for 31385c9
     angt

     authored
     31385c9View commit details

     Copy full SHA for 31385c9

     Browse repository at this point

   • ggml : collect all input tensors into graph_inputs (#29634)

     Show description for 8019dc5
     ggerganov

     authored
     8019dc5View commit details

     Copy full SHA for 8019dc5

     Browse repository at this point

   • ggml-zdnn: fix 0-row tensor crash (#29636)

     Show description for 86ea01d
     taronaeo

     authored
     86ea01dView commit details

     Copy full SHA for 86ea01d

     Browse repository at this point

   • musa: build the docker image and CI container from the MUSA SDK images (#29624)

     Show description for 18b74ff
     yeahdongcn

     authored
     18b74ffView commit details

     Copy full SHA for 18b74ff

     Browse repository at this point

   • ggml : speed up model loading (#29598)

     Show description for c13e04e
     

     angt

     and

     JohannesGaessler

     authored
     c13e04eView commit details

     Copy full SHA for c13e04e

     Browse repository at this point

   • ci: remove gpu-rocm keyed directory logs (#28940)

     Show description for c8cda8b
     taronaeo

     authored
     c8cda8bView commit details

     Copy full SHA for c8cda8b

     Browse repository at this point

   • llama : fix init in several tools/examples (#29632)

     ggerganov

     authored
     6d78fb0View commit details

     Copy full SHA for 6d78fb0

     Browse repository at this point

   • metal: FWHT perf optimizations (#29602)

     Show description for 18bbc46
     bri-prism

     authored
     18bbc46View commit details

     Copy full SHA for 18bbc46

     Browse repository at this point