Repository navigation

QVAC-21582 — Pull latest from upstream whisper.cpp (v1.9.1) - #73

Merged

freddy311082 merged 291 commits into

masterfrom

QVAC-21582/whisper-upstream-v1.9.1

Jun 30, 2026

Merged

QVAC-21582 — Pull latest from upstream whisper.cpp (v1.9.1)#73
freddy311082 merged 291 commits into
masterfrom
QVAC-21582/whisper-upstream-v1.9.1

This pull request is big! We're only showing the most recent 250 commits

Commits on May 14, 2026
- sync : ggml
  ggerganov
  committed
  4730e76
- talk-llama : sync llama.cpp
  ggerganov
  committed
  54ecc9d
- try to fix window cublas CI failure
  
  danbev
  authored and
  ggerganov
  committed
  f6f32a7
- Revert "try to fix window cublas CI failure"
  
  danbev
  authored and
  ggerganov
  committed
  1665885
- try using CCCL 12.4.127 with cuda 11.8.0 to fix CI failure
  
  danbev
  authored and
  ggerganov
  committed
  e0bfd3a
- Revert "try using CCCL 12.4.127 with cuda 11.8.0 to fix CI failure"
  
  danbev
  authored and
  ggerganov
  committed
  5b2d4af
- devops : add spirv-headers to vulkan dockerfile
  
  danbev
  authored and
  ggerganov
  committed
  633de7f
- ggml-cuda : add explicit casts to -INFINITY for float and half2 types
  
  danbev
  authored and
  ggerganov
  committed
  b1ebddf
- ggml-cuda : add ar_add() to avoid ambiguous operator+ for half/bfloat16 in CUDA 11.8
  
  danbev
  authored and
  ggerganov
  committed
  b6a4b32
- ci : update ONEAPI version to 2025.3.3-0-devel-ubuntu24.04
  
  danbev
  authored and
  ggerganov
  committed
  d04a1fa
- squash! ci : update ONEAPI version to 2025.3.3-0-devel-ubuntu24.04
  
  danbev
  authored and
  ggerganov
  committed
  ea29be5
- Revert "ggml-cuda : add ar_add() to avoid ambiguous operator+ for half/bfloat16 in CUDA 11.8"
  
  danbev
  authored and
  ggerganov
  committed
  db7bcdb
- Revert "ggml-cuda : add explicit casts to -INFINITY for float and half2 types"
  
  danbev
  authored and
  ggerganov
  committed
  5a24c75
- ggml: install ggml.pc in <libdir>/pkgconfig (ggml/1480)
  
  robUx4
  authored and
  ggerganov
  committed
  dd70679
- metal : tighten input-position loop in kernel_conv_transpose_1d (ggml/1477)
  
  CrispStrobe
  authored and
  ggerganov
  committed
  5f08683
- ggml-virtgpu : include missing mutex header (llama/22810)
  
  olliewalsh
  authored and
  ggerganov
  committed
  73f63f5
- Add OP im2col_3d (llama/22903)
  
  arthw
  authored and
  ggerganov
  committed
  4db2f45
- CUDA: directly include cuda/iterator (llama/22936)
  
  ORippler
  authored and
  ggerganov
  committed
  0077a6d
- vulkan: Support asymmetric FA in scalar/mmq/coopmat1 paths (llama/22589)
  
  jeffbolznv
  authored and
  ggerganov
  committed
  c0c1f99
- Ggml/cuda snake fusion hardening (llama/22912)
  
  ServeurpersoCom
  authored and
  ggerganov
  committed
  449b33f
- CUDA: handle OW > 65535 in im2col (2D and 3D) (llama/22944)
  
  CrispStrobe
  authored and
  ggerganov
  committed
  287f637
- opencl: add q4_1 MoE for Adreno (llama/22856)
  
  authored and
  ggerganov
  committed
  ea4652c
- metal : promote mul_mv/mul_mm batch divisors to function constants (llama/22711)
  
  guyfischman
  authored and
  ggerganov
  committed
  8ec91c9
- vulkan: Check shared memory size for mmq shaders (llama/22693)
  
  jeffbolznv
  authored and
  ggerganov
  committed
  20895ab
- vulkan: Fix Windows performance regression on Intel GPU BF16 workloads for Xe2 and newer (llama/22461)
  
  rillomas
  authored and
  ggerganov
  committed
  be5a35c
- ggml-webgpu: address precision issues for multimodal (llama/22808)
  
  Constannnnnt
  authored and
  ggerganov
  committed
  a9bcbf5
- ggml-webgpu: Enables running gpt-oss-20b (llama/22906)
  
  yomaytk
  authored and
  ggerganov
  committed
  e8a7cd3
- opencl: add opt-in Adreno xmem F16xF32 GEMM for prefill (llama/22755)
  
  authored and
  ggerganov
  committed
  1caed1d
- hexagon: eliminate scalar VTCM loads via HVX splat helpers (llama/22993)
  
  authored and
  ggerganov
  committed
  bcaf449
- ggml-zendnn : adaptive fallback to CPU backend for small batch sizes (llama/22681)
  
  z-sachin
  authored and
  ggerganov
  committed
  8b288f5
- hexagon: add unary tanh op (llama/22999)
  
  max-krasnyansky
  authored and
  ggerganov
  committed
  cb7d38b
- flush the gpu profile timestamp before the queryset is overflowed (llama/22995)
  
  yomaytk
  authored and
  ggerganov
  committed
  1cbbd0b
- opencl: fix crash when warming up MoE on Adreno (llama/22876)
  
  lhez
  authored and
  ggerganov
  committed
  b19beb6
- opencl: add q5_0 and q5_1 MoE for Adreno (llama/22985)
  
  authored and
  ggerganov
  committed
  d4a4d87
- Fix for issue #22974. Cast intermediate results to float before adding and casting the result to the destination type. Avoids half+half operator ambiguity. (llama/22994)
  
  scutler-nv
  authored and
  ggerganov
  committed
  97371e9
- ggml-webgpu: only use subgroup-matrix path when head dims are divisible by sg_mat_k / sg_mat_n (llama/23020)
  
  ArberSephirotheca
  authored and
  ggerganov
  committed
  e4ce42e
- sync : ggml
  ggerganov
  committed
  69500f5
- talk-llama : sync llama.cpp
  ggerganov
  committed
  46ca43d
Commits on May 15, 2026
- server: add support for carry_initial_prompt (#3781)
  alubbe
  authored
  968eebe
Commits on May 18, 2026
- server : Return speaker information in JSON (#3782)
  alubbe
  authored
  6227a0e
- examples : fix memory leak in read_audio_data (#3810)
  
  petterreinholdtsen
  and
  danbev
  authored
  47b9eb3
Commits on May 19, 2026
- whisper : set bench data for each iteration (#3812)
  danbev
  authored
  afa2ea5
Commits on May 21, 2026
- ci : use github ubuntu-22.04-arm runner instead of qemu (#3815)
  danbev
  authored
  8443cf0
Commits on May 22, 2026
- common : fix server /inference fails to decode in-memory audio (regression) (#3818)
  ServeurpersoCom
  authored
  0ccd896
Commits on May 25, 2026
- fix: in bindings/ruby/test/jfk_reader/jfk_reader in jfk_reader.c (#3756)
  
  OrbisAI Security
  and
  claude
  authored
  b3877e1
- cmake : add CMakePresets.json [no ci] (#3808)
  danbev
  authored
  e414ecf
- SYCL: fix multi-GPU system RAM exhaustion by using Level Zero allocations (llama/21597)
  
  authored and
  ggerganov
  committed
  06cfc36
- vulkan: fix matmul integer pipeline selection (llama/23005)
  
  0cc4m
  authored and
  ggerganov
  committed
  97ba443
- ggml-cpu: Add IME2 Instruction Support for the SpacemiT Backend (llama/22863)
  
  alex-spacemit
  authored and
  ggerganov
  committed
  f022390
- logs : reduce (llama/23021)
  ggerganov
  committed
  592a8cd
- ggml-webgpu: makes the flash attn vec path subgroup-aware (llama/23040)
  
  ArberSephirotheca
  authored and
  ggerganov
  committed
  13133ab
- HIP: RDNA3 mma FA, faster AMD transpose, tune AMD (llama/22880)
  
  JohannesGaessler
  authored and
  ggerganov
  committed
  e62d589
- ggml-hexagon: cpy: add contiguous fast-path in reshape copy (llama/23076)
  
  pdhinaka
  authored and
  ggerganov
  committed
  18a61f4
- llama + spec: MTP Support (llama/22673)
  
  am17an
  and
  ggerganov
  committed
  23f956d
- ggml : bump version to 0.12.0 (ggml/1494)
  ggerganov
  committed
  587dca0
- ggml-alloc: fix out-of-bounds read in ggml_dyn_tallocr_remove_block (ggml/1492)
  
  thesayakmondal
  authored and
  ggerganov
  committed
  3583e35
- ggml.h: correct ggml_silu_back arg docstring (a=dy, b=x) (ggml/1500)
  
  OriPekelman
  authored and
  ggerganov
  committed
  e78e693
- vulkan: removed duplicate #include <memory> in headers (llama/23144)
  
  winstonma
  authored and
  ggerganov
  committed
  ef5ddec
- vulkan: fuse SSM_CONV + BIAS + SILU (llama/22653)
  
  jeffbolznv
  authored and
  ggerganov
  committed
  c7dd64c
- vulkan: Support unaligned tensors for ROPE (llama/22637)
  
  jeffbolznv
  authored and
  ggerganov
  committed
  e417ce7
- vulkan: add cpy bf16 -> f32 pipelines (llama/22677)
  
  ServeurpersoCom
  authored and
  ggerganov
  committed
  50482cb
- ggml-vulkan/CMakeLists: add a check for SPIRV-Headers (llama/22009)
  
  jeeb
  authored and
  ggerganov
  committed
  9e96e0e
- CUDA: Continue directly including cuda/iterator (llama/23102)
  
  ORippler
  authored and
  ggerganov
  committed
  53736a3
- feat: Support d_conv=15 for ssm-conv.cu (llama/23017)
  
  gabe-l-hart
  authored and
  ggerganov
  committed
  4fb3cca
- sycl: route small f32 matmuls to oneMKL, bypass oneDNN (llama/22150)
  
  authored and
  ggerganov
  committed
  619262a
- sycl: scalar SWAR byte-subtract in Q6_K MMVQ dot product (llama/22156)
  
  authored and
  ggerganov
  committed
  c65b082
- ggml-hexagon: add PAD op HVX kernel (llama/23078)
  
  authored and
  ggerganov
  committed
  0a11c9f
- hexagon: add support for TRI op (llama/22822)
  
  authored and
  ggerganov
  committed
  eb558f2
- rpc : keep last_graph_uid in the device context (llama/23273)
  
  rgerganov
  authored and
  ggerganov
  committed
  3477fdb
- sycl: add GGML_SYCL_USE_ASYNC_MEM_OP env toggle (llama/22153)
  
  authored and
  ggerganov
  committed
  28edd0c
- ggml-webgpu : extend GDN for K>1 (llama/23299)
  
  reeselevine
  authored and
  ggerganov
  committed
  6090f39
- hexagon: enable support for NORM op (llama/23319)
  
  aparmp-quic
  authored and
  ggerganov
  committed
  459ff07
- hexagon: add MROPE and IMROPE support in HTP rope op (llama/23317)
  
  aparmp-quic
  authored and
  ggerganov
  committed
  aca63e7
- opencl: add MoE support for q4_k, q5_k, q6_k on Adreno (llama/23303)
  
  authored and
  ggerganov
  committed
  37f1720
- ggml-cuda: tune RDNA3 Q6_K MMVQ nwarps (llama/23349)
  
  ravel7524
  authored and
  ggerganov
  committed
  0a0a342
- metal : optimize pad + cpy (llama/23354)
  ggerganov
  committed
  c58fc46
- Programmatic Dependent Launch (PDL) for more performance on newer NVIDIA GPUs (Hopper+) (llama/22522)
  
  aendk
  authored and
  ggerganov
  committed
  3fa1955
- hexagon: HMX quantized matmul rework (llama/23368)
  
  authored and
  ggerganov
  committed
  b93a5ba
- vulkan: optimize operations in the IM2COL shader (llama/22685)
  
  daniandtheweb
  authored and
  ggerganov
  committed
  ad717a6
- opencl: refactor backend initilization (llama/23318)
  
  lhez
  authored and
  ggerganov
  committed
  896718e
- hexagon: ssm-conv fix for large prompts (llama/23307)
  
  authored and
  ggerganov
  committed
  6d1d66d
- ggml : Check the right iface method before using the fallback 2d get (llama/23306)
  
  TheBlueMatt
  authored and
  ggerganov
  committed
  03da9f1
- metal : optimize concat kernel and fix set kernel threads (llama/23411)
  ggerganov
  committed
  158d93c
- fix(flash-attn): replace f32 with kv_type and q_type (llama/23372)
  
  Constannnnnt
  authored and
  ggerganov
  committed
  c436f14
- vulkan: fuse snake activation (mul, sin, sqr, mul, add) (llama/22855)
  
  ServeurpersoCom
  authored and
  ggerganov
  committed
  8402c36
- CUDA: fix PDL CC check for JIT compilation (llama/23471)
  
  JohannesGaessler
  authored and
  ggerganov
  committed
  ec18355
- ggml-zendnn : add Q8_0 quantization support (llama/23414)
  
  z-sachin
  authored and
  ggerganov
  committed
  2d62953
- SYCL: add BF16 to DMMV kernel path (~4x tg speedup on Intel Arc) (llama/21580)
  
  authored and
  ggerganov
  committed
  6fb7f1a
- SYCL : gated_delta_net K>1 (llama/23174)
  
  karavayev
  authored and
  ggerganov
  committed
  0416fee
- sycl : Level Zero detection in ggml_sycl_init (llama/23097)
  
  sanmai
  authored and
  ggerganov
  committed
  21c65a7
- SYCL: improve MoE prefill throughput (llama/23142)
  
  sanmai
  authored and
  ggerganov
  committed
  b0c9f90
- opencl: generalize Adreno MoE kernels on M (llama/23449)
  
  shawngu-quic
  authored and
  ggerganov
  committed
  aefffa1
- vulkan: fix windows find_package of SPIRV-Headers (llama/23215)
  
  jeffbolznv
  authored and
  ggerganov
  committed
  6b85d73
- ggml : Check the right iface method before using the fallback 2d get (llama/23514)
  
  dskwe
  authored and
  ggerganov
  committed
  511f860
- hexagon: apply repl optimization in flash attn softmax as #22993 (llama/23455)
  
  njsyw1997
  authored and
  ggerganov
  committed
  b84d034
- opencl: batch profiling to improve speed and prevent memory leaks (llama/23495)
  
  shaofeiqi
  authored and
  ggerganov
  committed
  1435988
- TP: fix entirely zero-sized slices per device (llama/23525)
  
  JohannesGaessler
  authored and
  ggerganov
  committed
  3306af6
- ggml : Parallelize quant LUT init (llama/23595)
  
  jeffbolznv
  authored and
  ggerganov
  committed
  a369b39
- ggml : bump version to 0.12.1 (ggml/1508)
  ggerganov
  committed
  946d681
- sync : ggml
  ggerganov
  committed
  0a62a57
- talk-llama : sync llama.cpp
  ggerganov
  committed
  865ec17
- readme : add AMD ROCm/HIP GPU build instructions (#3823)
  Kaihui-AMD
  authored
  44a50ca
- ggml: `gguf_init_from_callback` and `gguf_init_from_buffer` (llama/22341)
  
  authored and
  ggerganov
  committed
  2979e5f
- TP: fix ggml context size calculation (llama/22616)
  
  JohannesGaessler
  authored and
  ggerganov
  committed
  bcff515
- ggml : bump version to 0.13.0 (ggml/1510)
  ggerganov
  committed
  1cf8e3a
- sync : ggml
  ggerganov
  committed
  f14ae77
- benches : update
  ggerganov
  committed
  c245b3e
- release : v1.8.5
  ggerganov
  committed
  e0fd1f6
Commits on May 26, 2026
- cli : merge tokens split across UTF-8 boundaries in JSON output (#3751)
  
  texasich
  and
  texasich
  authored
  27101c0
Commits on May 27, 2026
- docs : add AGENTS.md and CONTRIBUTING.md [no ci] (#3826)
  danbev
  authored
  ee540bf
- ci : only run docker jobs when pushed to master [no ci] (#3828)
  danbev
  authored
  6dcdd65
Commits on May 28, 2026
- ci : set GGML_NATIVE=OFF for bindings-java (#3830)
  danbev
  authored
  f6e617b
- ci : renable arm64 docker builds (#3832)
  danbev
  authored
  9186e24
- ci : add on push/pull_request paths ruby job (#3833)
  danbev
  authored
  f41562b
- ci : fix include paths for bindings-go job [no ci] (#3835)
  danbev
  authored
  e47a3ee
- ci : add ignore for bindings/{ruby, go} in build.yml [no ci] (#3837)
  danbev
  authored
  c932729
Commits on May 29, 2026
- CUDA: add fast walsh-hadamard transform (llama/23615)
  
  authored and
  ggerganov
  committed
  205ee5a
- metal : add apple device id (llama/23566)
  
  authored and
  ggerganov
  committed
  1c477d4
- CUDA: missing PDL sync for FWHT, better fallback (llama/23690)
  
  JohannesGaessler
  authored and
  ggerganov
  committed
  2307712
- Check batch_compute_passes before sending passes when not doing GPU profiling (llama/23457)
  
  nikhilJain17
  authored and
  ggerganov
  committed
  bc77933
- ggml-webgpu: Add MMVQ path for Q4/Q8/Q2_K/Q4_K and clean up legacy MUL_MAT pipeline (llama/23594)
  
  yomaytk
  authored and
  ggerganov
  committed
  00a5110
- SYCL: implement ggml_sycl_pool_vmm (llama/22862)
  
  authored and
  ggerganov
  committed
  049f0af
- hexagon: add support for CONCAT op (llama/23648)
  
  max-krasnyansky
  authored and
  ggerganov
  committed
  f8df28d
- vulkan: optimize conv2d and implement coopmat1 support (llama/22620)
  
  jeffbolznv
  authored and
  ggerganov
  committed
  a0efd13
- ggml-zendnn : fixed naming of matmul function (llama/20964)
  
  authored and
  ggerganov
  committed
  6a249cd
- vulkan: avoid preferring transfer queue on AMD UMA devices (llama/22455)
  
  winstonma
  authored and
  ggerganov
  committed
  80e87ec
- CUDA: restrict PDL to CTK >= 12.3 due to MSVC issues (llama/23742)
  
  ORippler
  authored and
  ggerganov
  committed
  98c6722
- vulkan: add REPEAT op support for f16 to f16. (llama/23298)
  
  l8bloom
  authored and
  ggerganov
  committed
  c5cde8c
- vulkan: use GL_NV_cooperative_matrix_decode_vector for faster matmul (llama/23541)
  
  jeffbolznv
  authored and
  ggerganov
  committed
  1b590bb
- vulkan: Switch MUL_MAT_VEC to 4 K per iteration for F16/32 (llama/22887)
  
  TheBlueMatt
  authored and
  ggerganov
  committed
  8bce478
- ggml-webgpu: Fix how to dispatch WG to some ops (llama/23750)
  
  yomaytk
  authored and
  ggerganov
  committed
  a52bd38
- hexagon: add support for Q4_1 in MUL_MAT and MUL_MAT_ID (llama/23647)
  
  authored and
  ggerganov
  committed
  3bbe933
- ggml-webgpu: remove legacy constants (llama/23672)
  
  reeselevine
  authored and
  ggerganov
  committed
  8c8f213
- opencl: OP_GATED_DELTA_NET (llama/23312)
  
  ymcki
  authored and
  ggerganov
  committed
  7e843a8
- Hexagon: OP_GATED_DELTA_NET K>1 support (llama/23531)
  
  ymcki
  authored and
  ggerganov
  committed
  d284e1c
- ggml: fixed Arm SVE usage bug in vec.h, vec.cpp (llama/22841)
  
  authored and
  ggerganov
  committed
  8e40325
- cuda : fix KQ mask offset integer overflow in fattn MMA kernel (llama/23610)
  
  authored and
  ggerganov
  committed
  60e420f
- vulkan: Fix memory logger unsafe iterator access (llama/23667)
  
  winstonma
  authored and
  ggerganov
  committed
  5db94ba
- vulkan: fix wrong index variable in inner loop (llama/23665)
  
  winstonma
  authored and
  ggerganov
  committed
  816c302
- vulkan: fast path for walsh-hadamard transform (llama/23687)
  
  jeffbolznv
  authored and
  ggerganov
  committed
  b896e91
- hexagon: minor refresh for HMX FA and MM (llama/23796)
  
  authored and
  ggerganov
  committed
  1b241b8
- CUDA: route batch>=4 quantized matmul to MMQ on AMD MFMA hardware (llama/23227)
  
  authored and
  ggerganov
  committed
  04795e6
- mmvq Optim: add MMVQ_PARAMETERS_TURING(mmvq_parameter_table_id) for … (#23729)
  
  authored and
  ggerganov
  committed
  4e8af44
- ggml: auto apply iGPU flag CUDA/HIP if integrated device (llama/23007)
  
  fl0rianr
  authored and
  ggerganov
  committed
  e1faa7c
- opencl: move backend info printing into its own function (llama/23702)
  
  lhez
  authored and
  ggerganov
  committed
  94922ce
- hexagon: basic/generic op fusion support and RMS_NORM+MUL fusion (llama/23835)
  
  max-krasnyansky
  authored and
  ggerganov
  committed
  442be17
- meta : Add missing `buffer` set in allreduce fallback !COMPUTE clear (llama/23480)
  
  TheBlueMatt
  authored and
  ggerganov
  committed
  f1b687d
- cuda : disables launch_fattn PDL enrollment due to compiler bug (llama/23825)
  
  aendk
  authored and
  ggerganov
  committed
  e90501e
- sync : ggml
  ggerganov
  committed
  cc65eb1
- talk-llama : sync llama.cpp
  ggerganov
  committed
  5828fba
- ggml : bump version to 0.13.1 (ggml/1523)
  ggerganov
  committed
  92fc3f2
- sync : ggml
  ggerganov
  committed
  f24588a
Commits on May 31, 2026
- common : re-implement `ffmpeg-transcode.cpp` + clarify ffmpeg usage (#3846)
  ggerganov
  authored
  f39cc71
- common : pass sample rate to `ffmpeg_decode_audio()`
  ggerganov
  committed
  6c343e7
- ci : remove obsolete self-hosted label
  ggerganov
  committed
  2e045a9
- pi : add config
  ggerganov
  committed
  099af1c
- ci : fix self-hosted paths to mnt
  ggerganov
  committed
  fe69461
Commits on Jun 1, 2026
- ci : fix path to whisper.h in examples.yml [no ci] (#3842)
  danbev
  authored
  0dff274
- release : v1.8.6
  ggerganov
  committed
  23ee035
Commits on Jun 2, 2026
Commits on Jun 4, 2026
- ci : refactor + optimize (#3847)
  
  ggerganov
  and
  danbev
  authored
  02d5316
- ci : only publish/push docker images daily (#3854)
  danbev
  authored
  12d1828
- ci : use ccache instead of sccache for windows-cublas [no ci] (#3855)
  danbev
  authored
  9302c06
- ci : pin github actions to commit SHAs (#3856)
  danbev
  authored
  7ecb08f
- ci : use emscripten-core and pin version (#3857)
  danbev
  authored
  ad17783
- ci: build-windows action slimming (#3858)
  danbev
  authored
  99613cb
Commits on Jun 6, 2026
- ci : add ccache to quantize, vad, and wasm jobs (#3860)
  danbev
  authored
  574fc0d
- ci : add HF_TOKEN to docker.yml workflow [no ci] (#3861)
  danbev
  authored
  a8ec021
Commits on Jun 8, 2026
- ci : add ccache to build-sycl [no ci] (#3859)
  danbev
  authored
  e1da83d
- model : support for DeepseekV32ForCausalLM with generic DeepSeek Sparse Attention (DSA) implementation (llama/23346)
  
  committed
  c50e951
- CUDA: Check PTX version on host side to guard PDL dispatch (llama/23530)
  
  authored and
  ggerganov
  committed
  f7aad4e
- ggml-webgpu: add q4_0/q8_0 SET_ROWS (llama/23760)
  
  reeselevine
  authored and
  ggerganov
  committed
  acd91d2
- ggml-webgpu: Check earlier for WebGPU required features (llama/23879)
  
  reeselevine
  authored and
  ggerganov
  committed
  9147a96
- vulkan: add Flash Attention support for BFloat16 KV cache (llama/23420)
  
  0cc4m
  authored and
  ggerganov
  committed
  4317ddb
- ggml : add some lsx support (llama/23798)
  
  MQ-mengqing
  authored and
  ggerganov
  committed
  64b0d6b
- metal : restore im2col implementation for large kernels (llama/23901)
  ggerganov
  committed
  bf74b55
- opencl: support bf16 by converting to f16 (llama/23839)
  
  lhez
  authored and
  ggerganov
  committed
  1c0d1f0
- sycl : Optimize Q3_K mul_mat by reorder (llama/23725)
  
  arthw
  authored and
  ggerganov
  committed
  687fbcb
- Add more types in GET_ROWS OP (llama/23710)
  
  arthw
  authored and
  ggerganov
  committed
  20323e4
- Support Q4_1, Q5_0, Q5_1 in Flash-attention (llama/23812)
  
  arthw
  authored and
  ggerganov
  committed
  ec0c661
- vulkan: Removed unused functions (llama/23175)
  
  winstonma
  authored and
  ggerganov
  committed
  aea93ad
- vulkan: Block-load Q3_K/Q6_K block data and subtract on 32b ints (llama/23056)
  
  TheBlueMatt
  authored and
  ggerganov
  committed
  982533f
- TP: quantized KV cache support (llama/23792)
  
  JohannesGaessler
  authored and
  ggerganov
  committed
  e815b26
- vulkan: reduce host memory lock contention (llama/23376)
  
  winstonma
  authored and
  ggerganov
  committed
  c471bcc
- vulkan: don't hold the device mutex while compiling pipelines (llama/23641)
  
  jeffbolznv
  authored and
  ggerganov
  committed
  71d80aa
- metal: template GLU kernels to support f16/f32 (llama/23882)
  
  shrivasshankar
  authored and
  ggerganov
  committed
  050b856
- opencl: add basic support for q5_0 and q5_1 (llama/23548)
  
  authored and
  ggerganov
  committed
  e728bae
- revert to using global_invocation_id for cpy shader (llama/23955)
  
  yomaytk
  authored and
  ggerganov
  committed
  db2a395
- opencl: fix compiler warnings for non-adreno path (llama/23922)
  
  lhez
  authored and
  ggerganov
  committed
  9a0265d
- clean up unused variables warnings (llama/23975)
  
  anavp-nvidia
  authored and
  ggerganov
  committed
  7922370
- hexagon: add gelu_quick (llama/24007)
  
  tboinovski1
  authored and
  ggerganov
  committed
  754247f
- hexagon: MUL_MAT, MUL_MAT_ID, FLASH_ATTN and GDN cleanup and optimizations for latest models (llama/23989)
  
  max-krasnyansky
  authored and
  ggerganov
  committed
  8d61a9e
- hexagon: profiler output fix and script updates (llama/24042)
  
  max-krasnyansky
  authored and
  ggerganov
  committed
  d31cb20
- opencl: use flat variants of q4_K and q6_K gemv for very large M (llama/24006)
  
  lhez
  authored and
  ggerganov
  committed
  f110ff5
- cuda: reserve space for quantize kv-cache at startup (llama/23907)
  
  authored and
  ggerganov
  committed
  d5a49eb
- ggml-cpu: use runtime SVE width in FWHT (llama/24059)
  
  chaxu01
  authored and
  ggerganov
  committed
  750fa4c
- Avoid PDL race conditions by disabling __restrict__ when PDL is used (llama/24030)
  
  authored and
  ggerganov
  committed
  00a9728
- ggml-cpu: extend RVV quantization vec dot to higher VLENs (llama/22754)
  
  authored and
  ggerganov
  committed
  a1a3186
- ggml-webgpu: FlashAttention refactor + standardize quantization support (llama/23834)
  
  reeselevine
  authored and
  ggerganov
  committed
  e9dbd0c
- metal : reduce rset heartbeat from 500ms -> 5ms (llama/24074)
  ggerganov
  committed
  9d6e561
- ggml: vectorize ggml_vec_dot_q4_1_q8_1 with WASM SIMD128 (llama/22209)
  
  sirohikartik
  authored and
  ggerganov
  committed
  991b5a8
- sycl : port multi-column MMVQ from CUDA backend (llama/21845)
  
  masonmilby
  authored and
  ggerganov
  committed
  4ecede8
- CUDA: enroll mul_mat_vec_q_moe into pdl (llama/24087)
  
  ORippler
  authored and
  ggerganov
  committed
  4fa1e06
- kleidiai : dynamic chunck-based scheduling for hybrid execution (llama/23819)
  
  chaxu01
  authored and
  ggerganov
  committed
  facb02c
- vulkan: add fwht support for Intel with shmem reduction (llama/23964)
  
  0cc4m
  authored and
  ggerganov
  committed
  5a1feed
- opencl: improve get_rows, cpy, concat and q6_k flat gemv (llama/24160)
  
  lhez
  authored and
  ggerganov
  committed
  a87e950
- vulkan: check coopmat2 features before reporting support (llama/24186)
  
  0cc4m
  authored and
  ggerganov
  committed
  1777def
- metal : fix im2col 1D case (audio models) (llama/24220)
  
  ngxson
  authored and
  ggerganov
  committed
  2c139c2
- HIP: add gfx1152 and gfx1153 to RDNA3.5 (llama/24129)
  
  harkgill-amd
  authored and
  ggerganov
  committed
  4669631
- sync : ggml
  ggerganov
  committed
  b932ec5
- ggml : bump version to 0.14.0 (ggml/1533)
  ggerganov
  committed
  b31466b
- sync : ggml
  ggerganov
  committed
  4df9a57
- talk-llama : sync llama.cpp
  ggerganov
  committed
  84bd03a
Commits on Jun 9, 2026
- coreml : fix --quantize crash for mlprogram format; fix --optimize-ane label (#3868)
  krystophny
  authored
  ba57392
- ci : pin github actions to commit sha's (#3865)
  danbev
  authored
  df7638d
Commits on Jun 15, 2026
- cuda: reset cuda context after reading memory size (llama/23935)
  
  0cc4m
  authored and
  ggerganov
  committed
  782f122
- vulkan: Use cm2 decode_vector for mul_mat_id B matrix loads (llama/23991)
  
  jeffbolznv
  authored and
  ggerganov
  committed
  fbf720d
- Implement 2D workgroups for scale, binary, and unary ops (llama/24044)
  
  nikhilJain17
  authored and
  ggerganov
  committed
  490e500
- Handle buffer overlap / buffer aliasing for concat operator (llama/24000)
  
  authored and
  ggerganov
  committed
  15e5d40
- ggml-webgpu: Improve prefill speeds for k-quants + refactor matmul for Q4/Q5/Q8 and k-quants (llama/24225)
  
  yomaytk
  authored and
  ggerganov
  committed
  aa42b48
- ggml-webgpu: Add clang-format job (llama/24308)
  
  reeselevine
  authored and
  ggerganov
  committed
  e69e513
- Remove case for GGML_TYPE_Q4_K in mvvq.cu (llama/23528)
  
  ravel7524
  authored and
  ggerganov
  committed
  72894aa
- ggml-cpu : fix rms_norm_back wrong output under in-place aliasing (llama/24305)
  
  devYRPauli
  and
  ggerganov
  committed
  2d68a30
- ggml : add GGML_OP_COL2IM_1D (llama/24206)
  
  ServeurpersoCom
  authored and
  ggerganov
  committed
  28c7ed3
- vulkan: add `v_dot2_f32_f16` support in matrix-matrix multiplication and Flash Attention (llama/24123)
  
  0cc4m
  authored and
  ggerganov
  committed
  686bc80
- vulkan: reduce iq1 shared memory usage for mul_mm (llama/24287)
  
  jeffbolznv
  authored and
  ggerganov
  committed
  dc79430
- CUDA: Fix ssm_scan_f32 data-races (llama/24360)
  
  ORippler
  authored and
  ggerganov
  committed
  ef85b26
- Remove padding and multiple D2D copies for MTP (llama/24086)
  
  gaugarg-nv
  authored and
  ggerganov
  committed
  1a1900f
- vulkan: use medium matmul tile on Asahi Linux (llama/24306)
  
  xingjianll
  authored and
  ggerganov
  committed
  a512e4c
- vulkan: add fast path for contiguous buffer transfers (llama/23973)
  
  winstonma
  authored and
  ggerganov
  committed
  6870cfd
- ggml : bump version to 0.15.0 (ggml/1539)
  ggerganov
  committed
  b04008f
- vulkan: ifdef eMesaHoneykrisp (build fix) (llama/24479)
  
  jeffbolznv
  authored and
  ggerganov
  committed
  afd5592
- opencl: add q5_0/q5_1 gemm and gemv kernels for Adreno (llama/24319)
  
  authored and
  ggerganov
  committed
  2dcfd49
- ggml: support concat for scalar types at cuda backend (llama/24011)
  
  zihaomu
  authored and
  ggerganov
  committed
  882736f
- ggml : bump version to 0.15.1 (ggml/1541)
  ggerganov
  committed
  f35f47b
- sync : ggml
  ggerganov
  committed
  0a3fa9c
- talk-llama : sync llama.cpp
  ggerganov
  committed
  0ec0845
Commits on Jun 16, 2026
- cli : add --version flag (#3878)
  rumitvn
  authored
  db5a84b
- release : v1.8.7 (#3881)
  danbev
  authored
  48f628a
- ci : only trigger release jobs for tags (#3883)
  danbev
  authored
  3805e60
- parakeet : add support for NVIDIA Parakeet (#3735)
  
  danbev
  and
  ggerganov
  authored
  9efddaf
Commits on Jun 17, 2026
- ruby : add support for Parakeet (#3885)
  KitaitiMakoto
  authored
  0d14756
- release : v1.9.0 (#3886)
  danbev
  authored
  86c40c3
Commits on Jun 18, 2026
- ci : add GGML_NATIVE=OFF and GGML_BMI2=OFF to windows-blas (#3891)
  danbev
  authored
  200b119
Commits on Jun 19, 2026
- release : v1.9.1 (#3892)
  danbev
  authored
  f049fff
Commits on Jun 29, 2026
- QVAC-21582 chore: sync vendored whisper.cpp + ggml to ggml-org upstream v1.9.1
  
  freddy311082
  and
  cursoragent
  committed
  25379c5
Commits on Jun 30, 2026
- QVAC-21582 chore: keep Tether .github, drop upstream CI workflows from merge
  
  freddy311082
  and
  cursoragent
  committed
  195e856
- Merge branch 'master' into QVAC-21582/whisper-upstream-v1.9.1
  freddy311082
  authored
  7d52b1d

Commits

Commits on May 14, 2026

Commits on May 15, 2026

Commits on May 18, 2026

Commits on May 19, 2026

Commits on May 21, 2026

Commits on May 22, 2026

Commits on May 25, 2026

Commits on May 26, 2026

Commits on May 27, 2026

Commits on May 28, 2026

Commits on May 29, 2026