Repository navigation

ggml: allow prefetching tensor overrides (#21067) - merge from master - #25

Open

crusaderky wants to merge 246 commits into

am17an:bunch-moe-transferfrom

crusaderky:bunch-moe-transfer-staging

Open

ggml: allow prefetching tensor overrides (#21067) - merge from master#25
crusaderky wants to merge 246 commits into
am17an:bunch-moe-transferfrom
crusaderky:bunch-moe-transfer-staging

Commits on Jul 27, 2026
- mtmd: support MiMo-V2.5 audio input (RVQ-based model) (#26190)
  Xuan-Son Nguyen
  authored
  1cbfd19
- Disable -ffast-math on HIP (#25495)
  Beinsezii
  authored
  91f8c9c
Commits on Jul 28, 2026
- contrib : add guideline about the "merge ready" label (#26178)
  ggerganov
  authored
  c6292cf
- spec: add eagle3-v3 support for gpt-oss model (#25794)
  ruixiang63
  authored
  f870678
- ggml-metal: FWHT kernel for metal backend (#25924)
  
  authored
  f95de97
- server : add extra trace log for prompt similarity (#26218)
  ggerganov
  authored
  9a3bf2b
- sycl: fix use-after-return of the SDPA scale in the oneDNN flash-attention path (#25880)
  
  meatposes
  and
  arthw
  authored
  d6b61ac
- common/chat: add specialized minimax m3 parser (#26210)
  aldehir
  authored
  6ba5ef2
- spec: add DSpark speculative decoding (#25173)
  
  authored
  8407527
- ggml-cuda: add chunked SSD matmul for Mamba-2 prefill acceleration (#22675)
  BLSharda
  authored
  b62b350
- vulkan: add iq4_nl support back to FA (#24585)
  jeffbolznv
  authored
  8161641
- ggml : set output of view src (#25729)
  
  authored
  da5b448
- server: abstract llama_memory calls to common_memory (#26221)
  Xuan-Son Nguyen
  authored
  ee3d1b5
- docs: Adapt conda-forge package name (#26229)
  
  jjerphan
  and
  dev-tinker
  authored
  ad77bd3
- ui: rendering performance follow-up (#26097)
  allozaur
  authored
  6e2bc65
- mtmd : add Nemotron 3 Nano Omni support (parakeet) (#22520)
  
  authored
  7e1e28c
- opencl: skip the Adreno KQ/KQV image kernels for multi-stream batches (#26189)
  wanghqc
  authored
  8190848
- ggml-webgpu: Fix some binding alias issues to support all archs, fix recurrent-state-rollback test (#25931)
  reeselevine
  authored
  bc71c24
- model: Add Laguna-S-2.1 LLM_TYPE (#26233)
  crusaderky
  authored
  e9fa078
Commits on Jul 29, 2026
Commits on Jul 30, 2026
- mimo2: address MTP review feedback (#26228)
  tnhnyzc
  authored
  64d528b
- [UT] enhance UT to show all real unsupported backends (#25234)
  
  arthw
  and
  ggerganov
  authored
  a6a77bc
- ggml : Fix issue with kleidiai ci and stringop overflow warning (#26277)
  JonathanC-ARM
  authored
  32703b4
- ui: IndexedDB and Conversations data fixes (#26278)
  allozaur
  authored
  21a5f5b
- metal: fix memory unwire if model is freed without any GPU operations (#26082)
  
  nikwen
  and
  forforever73
  authored
  d0bfb19
- Remove custom cpu op from the M3 graph, express with stock ops (#26297)
  timkhronos
  authored
  b2f2216
- CUDA: add Q2_0 support (#25707)
  khosravipasha
  authored
  9b2a088
- ggml : bump version to 0.18.0 (ggml/1576)
  ggerganov
  committed
  6a4c335
- sync : ggml
  ggerganov
  committed
  9ebfc3a
- readme : refresh (#26280)
  ggerganov
  authored
  6b36c23
- ggml-cuda: Allow transpose-free gemmv computation (#26171)
  roberteg16
  authored
  e1a1abb
- tests : avoid building get-model.cpp many times (#26317)
  ggerganov
  authored
  47f686f
- llama-context : sync pending async copies before clearing embd_seq (#25676)
  o7si
  authored
  432d7ff
- Test support for alternative conv layout (#25617)
  pmaybank
  authored
  958d9c0
- vulkan: Support quantized concat (#25684)
  jeffbolznv
  authored
  ea63b4d
- server: support inp embd to generate next token (#26313)
  Xuan-Son Nguyen
  authored
  b4ca032
- mtmd: add lanczos resize method [no release] (#26341)
  Xuan-Son Nguyen
  authored
  5f55650
Commits on Jul 31, 2026
- ggml-webgpu: improve flash_attn_vec for quantized KV at long contexts (#25956)
  yomaytk
  authored
  8f4646a
- sycl: fuse RMS_NORM + MUL (#26015)
  Titaniumtown
  authored
  1553725
- [SYCL] Support q2 mul_mat (#26231)
  arthw
  authored
  a2be61d
- sycl : support dev2dev memcpy by DEV2DEV_MEMCPY_FORWARD (#26234)
  
  arthw
  and
  NeoZhangJianyu
  authored
  1c5b89f
- ggml-zendnn : group matmul direct API for mul_mat_id (#25918)
  z-sachin
  authored
  1e22599
- llama : enforce the same K and V cache types for DeepSeek V4; enable FA if V cache is quantized (#25871)
  
  fairydreaming
  and
  sszymczy
  authored
  69e62fc
- [SYCL] support the missed types in cpy (#26005)
  arthw
  authored
  d5d3e05
- SYCL: add oneMKL GEMM flash attention for XMX-accelerated prompt proc… (#25025)
  
  authored
  9d9a6d2
- cuda: extract Q2_0 elements via __byte_perm (#25603)
  dfriehs
  authored
  15e755f
- server: correct accepted tokens when need draft token replay (#26320)
  
  ruixiang63
  and
  ggerganov
  authored
  0005475
- vulkan: update vulkan sdk to 1.4.357.0 (#26303)
  jeffbolznv
  authored
  6f3c0a7
- llama : load MTP tensors only if they are really used (#26296)
  
  fairydreaming
  and
  sszymczy
  authored
  82dbc4f
- Support rotated kv cache quant (#26180)
  timkhronos
  authored
  a09d8ab
- mtmd: add n_embd_head (#26342)
  
  Xuan-Son Nguyen
  and
  shimmyshimmer
  authored
  db7d8b2
- vulkan: Introduce driver version check for Windows Intel GPU to mitigate crashing (#25192)
  rillomas
  authored
  eb41d50
- vulkan: add POOL_1D op (#25431)
  AnandPatil1
  authored
  876a432
Commits on Aug 1, 2026
- chat : enable tool call in thinking for DS4 (#26269)
  pwilkin
  authored
  ddd4ec1
- mtmd: add minicpmv46 downsample (#25993)
  tc-mb
  authored
  de69995
- cli : persist reasoning_content in chat history (#26362)
  ramicaza
  authored
  c629da5
- agents: clarify comment style and jinja knowledge (#26405)
  Xuan-Son Nguyen
  authored
  89482bd
- vendor : update BoringSSL to 0.20260730.0 (#26353)
  cabelo
  authored
  815a2a5
- vulkan: extend topk_moe fusion to support sqrt(softplus) (#26124)
  jeffbolznv
  authored
  a7a6d0d
- test: fix some CI errors (#26415)
  Xuan-Son Nguyen
  authored
  11924d4
Commits on Aug 2, 2026
- ggml-webgpu: add support for f16 repeat (#26307)
  yomaytk
  authored
  7a2db1a
- model : load MiMo V2 MTP tensors only if used (#26412)
  CISC
  authored
  75587a0
- sycl: fix classification of iGPUs (#26105)
  KyleHagy
  authored
  272700b
- chat : add qwen3 specialized parser (#26252)
  
  aldehir
  and
  bartdeboer
  authored
  f5919bf
- DeepseekV4 MTP + DSpark (#25784)
  am17an
  authored
  596a579
- opencl: bugfix increment ref_count in ggml_backend_opencl_init() (#26162)
  akleine
  authored
  c745be2
- convert: add option to create separate dspark GGUF (#26452)
  am17an
  authored
  3581ba0
- common: support the DSpark sidecar resolution (#26458)
  ServeurpersoCom
  authored
  bb4e0e1
- metal: implement DeepSeek V4 hyper-connections (#26459)
  
  ggerganov
  and
  tarruda
  authored
  fffbcbd
- opencl: limit local workgroup size for GLU operation (#26383)
  mgroeber9110
  authored
  0ab9d6f
- metal : add F16 support for bin ops (#26465)
  ggerganov
  authored
  9d21b57
- metal : add SILU_BACK (#25982)
  Blackcyan30
  authored
  221f0f6
Commits on Aug 3, 2026
- metal: implement DSv4 Lightning Indexer (#25893)
  
  authored
  1464c62
- llama : MTP support for DeepSeek V3.2 (#26457)
  
  fairydreaming
  and
  sszymczy
  authored
  2b63e06
- model: MTP support for Qwen3-Next (#25589)
  
  yomaytk
  and
  CISC
  authored
  4ed2b13
- server: (tools) add x-tool-cwd header (#26420)
  Xuan-Son Nguyen
  authored
  f2b52a8
- server: add notice for upcoming default port change 8080 --> 9931 (#26508)
  Xuan-Son Nguyen
  authored
  0b14b87
- CUDA: Fix data-races when reusing SMEM in block_reduce (#26385)
  ORippler
  authored
  9bd4c09
- CUDA: Add backend sampler for penalties sampler (#25262)
  kmorennv
  authored
  96278e3
- llama : allocate indexer cache only in "full" indexer layers (#26474)
  
  fairydreaming
  and
  sszymczy
  authored
  563dec8
- model: M3: Move MSA into a new memory implementation (#26338)
  
  timkhronos
  and
  ggerganov
  authored
  67d5978
- graph : fix unused input tensors in minimax m3 graph (#26519)
  ggerganov
  authored
  c50b34a
- opencl: route large q6_K lm_head to the flat GEMV (#26427)
  wanghqc
  authored
  39eab74
- ggml: use dynamic allocation for split graph inputs (#22789)
  
  AgoraPete
  and
  ggerganov
  authored
  dbadb68
- vocab : validate default special token ids (#26506)
  CISC
  authored
  e8e06f7
- server: add get_info tool (#26522)
  
  Xuan-Son Nguyen
  and
  ServeurpersoCom
  authored
  99111b1
- tests: add model resolution test on synthetic repo listings (#26172)
  
  ServeurpersoCom
  and
  Xuan Son Nguyen
  authored
  ee0445c
- model : support MTP in GLM-4.7-Flash (#24868)
  jacekpoplawski
  authored
  57c0921
- vendor : update BoringSSL to 0.20260803.0 (#26523)
  cabelo
  authored
  fe2adf0
- vendor : update cpp-httplib to 0.52.0 (#26485)
  cabelo
  authored
  94bc47f
- chat : add new template for DeepSeek V4 Flash 0731 (#26398)
  tarruda
  authored
  0ef6e55
Commits on Aug 4, 2026
- Extended SYCL oneDNN SDPA to non-FP16 KV caches (Q4_0–Q8_0 and FP32) (#25874)
  
  johnkarlhill
  and
  claude
  authored
  66fa168
- sycl: parallelize the non-contiguous concat kernel (#25852)
  Titaniumtown
  authored
  6c8dcaa
- ci: fix vulkan llvmpipe runs (#26533)
  netrunnereve
  authored
  22dc605
- llama : move n_vocab from llama_sampler_data to penalty_sampler (#26520)
  ggerganov
  authored
  935cad6
- model : allow reshape of tensors during load (#26531)
  ggerganov
  authored
  1269cb1
- convert : import bytes_to_unicode from convert_slow_tokenizer (#26217)
  SolshineCode
  authored
  b06fbc9
- vocab : validate plamo2 byte tokens (#26511)
  CISC
  authored
  cf06ad7
- vulkan backend ops: implemented GATED_LINEAR_ATTN (#25601)
  PranavUttarkar
  authored
  f26efa0
- convert : add missing return after setting tekken vocab (#25947)
  AngelGalindo7
  authored
  b5746d2
- ggml : bump version to 0.18.1 (ggml/1578)
  ggerganov
  committed
  15831f5
- sync : ggml
  ggerganov
  committed
  2e17f69
- gguf-py: validate n_dims and guard against uint64 overflow in reader (#25401)
  hclsys
  authored
  5788b51
- speculative : refactor enabled configs common_speculative_init (#26510)
  danbev
  authored
  7bd8282
- ci: fix pre-built binaries no longer working on macOS 15 and below (#26375)
  nikwen
  authored
  6b5224c
- models : fix dflash wo_a reshape on load (#26577)
  ggerganov
  authored
  1c3c967
- mtmd: support Qwen3-TTS (note: breaking change to llama-tts binary) (#26254)
  
  Xuan-Son Nguyen
  and
  ServeurpersoCom
  authored
  0713275
- ui: CWD for agent (#26518)
  
  allozaur
  and
  ServeurpersoCom
  authored
  2f56fc3
- gguf-split: Add option to delete split parts during merge (#26538)
  
  guiq83
  and
  CISC
  authored
  76c956c
- sampler : remove "full-context windows" from history-based samplers (#26524)
  ggerganov
  authored
  a6aa6f5
- mtmd: correcting duplicate empty audio chunks for short inputs (#26536)
  Abhinaykrishna2
  authored
  474c92e
- server: decode Windows OEM output to UTF-8 in built-in tools (#26597)
  ServeurpersoCom
  authored
  4308a4f
- Prefer npm ci over install for security (#26601)
  eapache
  authored
  6ea215d
Commits on Aug 5, 2026
- ui: Update vulnerable packages + cleanup Storybook config (#26607)
  allozaur
  authored
  e031d95
- build : remove GGML_METAL_USE_BF16 from all build scripts (#26604)
  nikwen
  authored
  d52ec04
- ui: show generation statistics by default in chat settings (#26624)
  allozaur
  authored
  3e3a7a4
- vendor : apply patches for subprocess.h (#26606)
  Xuan-Son Nguyen
  authored
  61881b1
- convert: Add endianness conversion for Q1 and TQ2 quantizations (#26618)
  
  Andreas-Krebbel
  and
  CISC
  authored
  020760a
- server: Adding spec-decode counters to /metrics endpoint (#26389)
  BLSharda
  authored
  a035a88
- security : clarify about AI-generated reports (#26579)
  Xuan-Son Nguyen
  authored
  9303cdd
- fit: Fix memory allocation for MTP layers (#26605)
  smalinin
  authored
  9a688e5
- mtmd: support multi-row batching for deepseek-ocr (#26154)
  
  Xuan-Son Nguyen
  and
  sfallah
  authored
  717dad5
- grammar : degrade max repetition >= 2000 to unbounded (#26613)
  aldehir
  authored
  cd0fa60
- mtmd: Unlimited-OCR fix max_tiles, setting in converter (#25614)
  sfallah
  authored
  b06aa77
- tests: re-enable MiniMax M3 in `test-llama-archs` (#26633)
  nikwen
  authored
  360e134
- server: harden the file_glob_search directory walk (#26626)
  ServeurpersoCom
  authored
  f9e832c
- mtmd/ggml: add ggml_build_forward_order (#26649)
  ServeurpersoCom
  authored
  c8e03ce
Commits on Aug 6, 2026
- vulkan: fix submission batching size, add debug tools for diagnosing causes of DeviceLost drivers errors (#26371)
  0cc4m
  authored
  803b7fc
- model-conversion : add --model-name to conversion scripts (#26665)
  danbev
  authored
  9de0fcf
- ci : onboard AMD ROCm CI with gfx1151 fixes (#26544)
  
  authored
  a1f96d4
- convert : accept "ExaoneMoeForCausalLM" arch spelling (#26660)
  csabakecskemeti
  authored
  e700bfb
- model-loader : fix quantized reshaped tensor strides (#26672)
  ggerganov
  authored
  3db4ff8
- convert : fix DeepseekV4 rope parameters with transformers 5.x (#26673)
  CISC
  authored
  eb5667a
- server: fix empty response for /cors-proxy (#26656)
  Xuan-Son Nguyen
  authored
  6a32c29
- mtmd: add chunk save/load function (#26645)
  Xuan-Son Nguyen
  authored
  15586e2
Commits on Aug 7, 2026
- metal : avoid `threadgroup` matrix array instantiation in kernel_lightning_indexer (#26646)
  JamePeng
  authored
  e40bf88
- ci: abort if build requirements are missing (#26368)
  nikwen
  authored
  d8d9887
- cuda: fix warnings for unused variable/function (#26688)
  dfriehs
  authored
  5b87ed3
- sycl : enhance OP set_rows to support all missed data types (#26515)
  arthw
  authored
  c074cb3
- sycl : fix error Error OP FLASH_ATTN_EXT on arc770 (#26441)
  arthw
  authored
  eef5f3e
- sycl : update guide Q&A and script for device setting (#26442)
  arthw
  authored
  c1f4109
- sycl : Support DSv4 OPs: LIGHTNING_INDEXER,DSV4_HC_COMB,DSV4_HC_POST,DSV4_HC_PRE (#26568)
  arthw
  authored
  31558db
- sycl: *glu flat path (#26354)
  Titaniumtown
  authored
  6b5c2ef
- sycl: fix UE4M3 parsing (#25608)
  clee
  authored
  fc3f10b
- Mitigate crashing issue on Windows MSYS2 UCRT64 environment (GCC 16.1.0) (#26555)
  rillomas
  authored
  42e9881
- ui: read model modalities from the router model list (#26709)
  ServeurpersoCom
  authored
  82bb485
- ggml : add aarch64 HWCAP fallbacks and fix fp16 variant detection (#25554)
  kilian-hu
  authored
  cb26014
- mtmd: stop feeding the text stream again during Qwen3-TTS generation (#26706)
  ServeurpersoCom
  authored
  217df17
- server: (router) do not evict busy models (#26567)
  Xuan-Son Nguyen
  authored
  e1470ee
- server: (router) add LRU scheduler (#26572)
  Xuan-Son Nguyen
  authored
  dff15d4
- ui: set npm `min-release-age` to protect against supply-chain attacks (#26711)
  nikwen
  authored
  34e9ee5
- webui: load the model selected via ?model= when ?load=true (#26707)
  erusev
  authored
  f4f7758
- server : clarify comment in eval_llama_cmpl_schema [no ci] [no release] (#26720)
  danbev
  authored
  9ba73fd
- ggml : bump version to 0.19.0 (ggml/1581)
  ggerganov
  committed
  933f46f
- sync : ggml
  ggerganov
  committed
  4cf5cab
- mtmd: fix longest_edge ignoring min/max pixels (#26638)
  Xuan-Son Nguyen
  authored
  4cb22cd
- ui: Filesystem `@mentions` for Chat Form (#26715)
  ServeurpersoCom
  authored
  2363478
- metal : fix NORM/RMS_NORM for row lengths that leave a partial simdgroup (#26708)
  robertomeroni
  authored
  a194a75
- sycl: coalesce the ssm_conv window loads (#26612)
  Titaniumtown
  authored
  f8e3026
- allozaur/feat/chat slash commands (#26716)
  ServeurpersoCom
  authored
  6de1b63
- tests : speed-up server test suite 3x (#26734)
  ggerganov
  authored
  1621a3d
- allozaur/feat/chat form contenteditable (#26717)
  ServeurpersoCom
  authored
  fc6545d
- tts: account for the vocoder pass in the timings line (#26733)
  ServeurpersoCom
  authored
  3653e6d
Commits on Aug 8, 2026
- CUDA: fix thread/block count in quantized cpy kernel launches (#26731)
  grafail
  authored
  69bf643
- server: add initial tool isolation support (via docker) (#26507)
  
  Xuan-Son Nguyen
  and
  ServeurpersoCom
  authored
  dd2c7c4
- server, ui: only offer a working directory when a tool reads it (#26762)
  ServeurpersoCom
  authored
  18f7ad7
- CUDA: fuse rms_norm + mul + rope (+ view + set_rows) (#26767)
  grafail
  authored
  687e778
- server: report the isolate working directory from get_info (#26773)
  
  ServeurpersoCom
  and
  Xuan-Son Nguyen
  authored
  7ba604f
Commits on Aug 9, 2026
- ci: rm `GGML_HIP_ROCWMMA_FATTN` (#26760)
  taronaeo
  authored
  61141f1
- ggml-cpu : fix missing Q5_0 dispatch in SpaceMiT backend (#26792)
  Hao-Chen2337
  authored
  0865990
- ci: add pr-draft-label (#26801)
  Xuan-Son Nguyen
  authored
  9369185
- ui: degrade the working directory picker when file search is off (#26811)
  ServeurpersoCom
  authored
  74ce157
Commits on Aug 10, 2026
- ggml-webgpu : refactor several wgsl files and simplify flash_attn wgsl. (#26134)
  yomaytk
  authored
  f401bb1
- ci: fix the ctest sanitize runs (#26593)
  netrunnereve
  authored
  aea252f
- model-saver : fix expert shared/chunk FFN length key clobber (#26693)
  SolshineCode
  authored
  0377426
- server: gate the docker tools runtime tests on a real container run (#26826)
  ServeurpersoCom
  authored
  1e396e7
- ui: Linting & Formatting scripts (#26819)
  allozaur
  authored
  92d1bb0
- readme : remove dev branches (#26832)
  ggerganov
  authored
  6ad4ab0
- model : Granite-Switch Architecture (#25107)
  barvhaim
  authored
  157b81f
- vendor : update cpp-httplib to 0.53.0 (#26821)
  cabelo
  authored
  e23e944
- model: add MTP support for Nemotron model (#26725)
  ruixiang63
  authored
  7a20b41
- ci: Add support for CUDA 13.4 ARM64 builds for Windows (#26650)
  
  authored
  2e2d99c
- llama: Restore quantization of mmprojs (#26818)
  
  pcuenca
  and
  Xuan Son Nguyen
  authored
  86c298f
- vendor: sync subprocess.h and drop local patches (#26808)
  ServeurpersoCom
  authored
  4c6766f
- chat : Align Laguna-S-2.1 chat template to huggingface (#26232)
  crusaderky
  authored
  a52077c
- model: Muse Glimmer Support (#26841)
  
  authored
  62bf73d
- server: add more tool isolation support (ssh remote + podman rootless) (#26774)
  
  ServeurpersoCom
  and
  Xuan Son Nguyen
  authored
  4ae84de
- ci : don't specify python version in server-sanitize for broader runner compatibility (#26840)
  CISC
  authored
  e5275f6
- ui: UI/chat form follow ups (#26743)
  ServeurpersoCom
  authored
  4dee52f
- ggml : require contiguous src for ROLL on CUDA and Metal (#25928)
  devYRPauli
  authored
  f8def7f
- ggml-cpu : fix CPU affinity mask being ignored on Android (#26838)
  hiteshchopra11
  authored
  d2f8305
- llama : support multi-output backend sampling (#25532)
  
  gaugarg-nv
  and
  ggerganov
  authored
  dd1ea52
- ci : target ROCm 7.14 for build and release (#25775)
  superm1
  authored
  0666ad2
- opencl: transpose the K tile in local memory for FA prefill kernels (#26428)
  wanghqc
  authored
  689e227
- Address review comment of PR 25532 (#26852)
  gaugarg-nv
  authored
  030ebb5
Commits on Aug 11, 2026
- ggml-webgpu: fix CI errors from #25025 and #25262 (#26566)
  yomaytk
  authored
  84f7129
- common/peg : suppress incomplete escape sequences (#26780)
  aldehir
  authored
  48d22e2
- model : fix SWA not being enabled for EXAONE 4.5 (#26848)
  junmo-kim
  authored
  14e78dd
- tests : disable backend sampler hip multi output (#26878)
  
  jimw567
  and
  Jim Wu
  authored
  4801e3c
- tests : clean-up server test, use `tests.sh` in ci (#26886)
  ggerganov
  authored
  b3df572
- llama: add default load-mode auto, which avoids mmap on iGPUs (#26081)
  
  authored
  153d324
- tests : fix running server tests on windows (#26889)
  ggerganov
  authored
  9afff1b
- model-conversion : use save_output_data for causual embeddings [no ci] (#26890)
  danbev
  authored
  1138b85
- ci: hip-quality-check: update vgpr spill ignore list (#26859)
  IMbackK
  authored
  7044859
- ui: fix context gauge for single-model usage (#25738)
  intel00000
  authored
  8d274dd
- mtmd: support pocket-tts (#26871)
  
  Xuan-Son Nguyen
  and
  ServeurpersoCom
  authored
  6e62ba5
- Dflash support for nemotron-3.5 (#26905)
  
  lnigam
  and
  CISC
  authored
  cc078b4
- convert : keep quantization scales for nemotron --mtp export (#26903)
  ynankani
  authored
  5d16e81
- requirements: use stable torch packages on s390x (#26864)
  nikwen
  authored
  2468576
- imatrix.cpp: Move finite check and only check touched experts (#26861)
  bartowski1182
  authored
  38406d5
- ci : add windows-rocm to check-release (#26897)
  CISC
  authored
  70dfba5
- chat : tighten bare function parsing for Qwen models (#26793)
  aldehir
  authored
  ba360ef
- spec : update speculative-simple (#26904)
  ggerganov
  authored
  f785fc9
- cuda : add warp-per-row wkv7 kernel for single-token decode (#26111)
  123123213weqw
  authored
  5988633
- CUDA: only disable CUDA graphs when mul_mat_id actually needs a stream sync (#26802)
  grafail
  authored
  ebb546b
- ci : add missing release check (#26923)
  CISC
  authored
  7b13a84
- chat : fix muse-glimmer detection of tool calls after EOM (#26879)
  ruanslv
  authored
  0b1bad1
Commits on Aug 12, 2026
- opencl: use flat mv q5_k when weight exceeds image1d_buffer_t limit (#26880)
  lhez
  authored
  cb27fe9
- convert : handle per_layer_config in Gemma4 (transformers 5.15) (#26882)
  
  pluvium27
  and
  CISC
  authored
  6eff593
- wavtokenizer-dec : bound posnet/convnext block_count against n_layer_all (#26892)
  
  oakkaya
  and
  CISC
  authored
  55f453b
- vulkan: add TQ2_0 (ternary) support (#25850)
  
  authored
  a7cd2f0
- tests : update speculative params (#26925)
  ggerganov
  authored
  a4a4c51
- opencl: default FA c8 cluster width to 16 on X1E (#26433)
  wanghqc
  authored
  89e0aa6
- ui: add read_media tool (#25877)
  
  authored
  4dd1275
- server : support slot save/restore with media inputs (#26640)
  CHIPMUNK-T0T
  authored
  5d9e5ac
- cmake : add config version support (ggml/1582)
  
  danbev
  authored and
  ggerganov
  committed
  13fd0bb
- sync : ggml
  ggerganov
  committed
  af05a42
- model : disallow integer dflash sliding_window_pattern (#26900)
  CISC
  authored
  ece98b8
- kleidiai: Add runtime feature detection mechanism for aarch64/kleidiai (#26076)
  JonathanC-ARM
  authored
  132753b
- gguf : harden loader against malformed tensor dims and metadata types (#25596)
  
  harrison001
  and
  ggerganov
  authored
  d8a8bea
- cmake : introduce semantic versioning (#26839)
  
  danbev
  and
  ggerganov
  authored
  680a9ae
- disable rocm cache (#26962)
  ggerganov
  authored
  7a9ff95
- ci : disable ubuntu-rocm (#26969)
  ggerganov
  authored
  9558fa4
- ci: fix thread sanitizer + remove ccache (#26927)
  
  netrunnereve
  and
  CISC
  authored
  84e908c
Commits on Aug 13, 2026
- Merge beellama-llaamacpp-base into bunch-moe-transfer
  crusaderky
  committed
  2d1a4e8
- Fix merge errors
  crusaderky
  committed
  0efd589
- cosmetic
  crusaderky
  committed
  44e6fb2

Commits

Commits on Jul 27, 2026

Commits on Jul 28, 2026

Commits on Jul 29, 2026

Commits on Jul 30, 2026

Commits on Jul 31, 2026

Commits on Aug 1, 2026

Commits on Aug 2, 2026

Commits on Aug 3, 2026

Commits on Aug 4, 2026

Commits on Aug 5, 2026