Repository navigation

Merge upstream master (d854d98ba) into prism - #352

Merged

bri-prism merged 1290 commits into

prismfrom

merge/upstream-20261011

Oct 30, 2026

Merged

Merge upstream master (d854d98ba) into prism#352
bri-prism merged 1290 commits into
prismfrom
merge/upstream-20261011

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

Commits on Oct 15, 2026
- opencl: fix OOB tail access for adreno noshuffle gemv kernels (#30421)
  
  lhez
  and
  wanghqc
  authored
  9166cea
- sycl : replace oneAPI package by DLE+oneDNN package (#30131)
  arthw
  authored
  e36df60
- vulkan : fix concurrent first device init (#29766)
  losewayy
  authored
  47f5a64
- jinja : write nan and inf in tojson as json.dumps does (#30483)
  ykhrustalev
  authored
  82e1e2f
- jinja : improve context copy constructor performance (#30562)
  Ninja3047
  authored
  08698c8
- models : add maion-coder (#29778)
  
  mailkv23
  and
  Karthik Vijayan
  authored
  3e4dfa8
- fix(ui): take the agentic turns default from DEFAULT_AGENTIC_CONFIG (#30582)
  allozaur
  authored
  fd55cb9
- ci : bump ty to 0.0.88 (#30587)
  CISC
  authored
  711780f
- ggml : bump version to 0.27.0
  ggerganov
  committed
  4ed1c85
- sync : ggml
  ggerganov
  committed
  c0cb53a
- server: prefer empty slots and keep states overwritten by child tasks (#30567)
  ServeurpersoCom
  authored
  b0d2391
- server: save the draft state in the slot file (#30574)
  ServeurpersoCom
  authored
  ea736c1
- model : Mellum - read MoE gating params from GGUF, unify RoPE (#30578)
  Xarbirus
  authored
  a6b91d8
- server: fix the off-by-two in the SWA reuse threshold (#30580)
  ServeurpersoCom
  authored
  61ad80b
- scripts : add ci-status.sh to list active CI runs (#30586)
  ggerganov
  authored
  a11b653
- models : add Nemotron Super 3.5 VL and Nemotron Bolt 74B A9B support (#30520)
  frozenblade1224
  authored
  cd0a5b6
- llama.cpp : bump version to 0.7.0 (#30603)
  ggerganov
  authored
  620d222
- pi : expand AI disclosure with hardware and stats (#30615)
  ggerganov
  authored
  b76b133
- model: add Muse Spark Open (text, vision and audio) (#30616)
  
  authored
  ba3c05f
- mtmd: lightonocr does not require longest edge to always be 1540, native resolution is the default (#30600)
  staghado
  authored
  2f729cf
- convert : fix regression in mapping of Granite speech tensors (#30628)
  mrutkows
  authored
  e287bd8
- convert : fix Granite Switch block_count for MultiSwitch checkpoints (#30621)
  
  planetf1
  and
  gabe-l-hart
  authored
  59d5fad
- mtmd: add mtmd_audio_stream + support voxtral-realtime (#30629)
  ngxson
  authored
  2851b3b
Commits on Oct 16, 2026
- [SYCL] Enable ssl in build script and CI (#30456)
  
  arthw
  and
  CISC
  authored
  43d57fb
- sycl: support reordered Q8_0 MoE weights (#29337)
  newjordan
  authored
  e707790
- Limit max CUDA pinned host buffer size to 4 GB (#30449)
  gaugarg-nv
  authored
  b6334e3
- llama : share the k-pool input views between layers (#30613)
  am17an
  authored
  6afa9a3
- hexagon: tensor repacking rework and support for tiled and flat quantized and fp weights (#30635)
  max-krasnyansky
  authored
  b797b9c
- graph : add precision handling for MELLUM architecture (#30584)
  Xarbirus
  authored
  5e03a1f
- Fix failing RocM tests (#30611)
  
  mailkv23
  and
  Karthik Vijayan
  authored
  f12aaa3
- vulkan: fix gpu-visible host memory issues (#30468)
  0cc4m
  authored
  a8d9a7e
- tests : catch vanished rows in the fusion baseline (#30618)
  ggerganov
  authored
  88aaf32
- sycl: wide stores for the q4_K/q5_K weight dequant (#29696)
  Titaniumtown
  authored
  13de11e
- sycl : reduce the build threads to half (#30236)
  arthw
  authored
  c0083e3
- ggml-cpu: refactor s309x repack helper functions (#30588)
  taronaeo
  authored
  c804754
- cpu : reduce AVX2 tiled accumulators together (#30296)
  trivalik
  authored
  3dbcb7f
- convert : support text-only checkpoints in pplx-decider (#30575)
  final-bench
  authored
  0f43762
- common : read ngram cache parts by reference instead of copying (#30061)
  jadidbourbaki
  authored
  13b8048
- CUDA: HIP: rms_norm: add variants that load float4 at a time and store x in registers (#29720)
  IMbackK
  authored
  623c163
- gguf-py : add numpy Q1_0 and Q2_0 de/quantization support (#30630)
  SomeNobody21112
  authored
  495c580
- [SYCL]: fix memory errors in mul_mat, split buffer, host pool (#29889)
  mctylr-gh
  authored
  a30fb49
- convert : speed up model endian conversion and add more datatype support (#30647)
  taronaeo
  authored
  b84ea5c
- model : MTP support for Mellum2.1 (#30612)
  Xarbirus
  authored
  9a16f1d
- hexagon: correct HMX FA sinks (#30581)
  
  jhen0409
  and
  max-krasnyansky
  authored
  ceefd16
- Add more benchmarks to llama-eval (#26174)
  pwilkin
  authored
  f336359
- convert : respect Mistral format for DeepseekV2Model (#30669)
  aiqubits
  authored
  2a66703
- common: honor HTTP_PROXY/HTTPS_PROXY env vars in http client (#21752)
  
  authored
  7182ef7
- model : Granite Switch Decision (#30676)
  gabe-l-hart
  authored
  23f7aab
- model: add Beam (Reflection) (#30670)
  danielhanchen
  authored
  01fc3aa
- opencl: Add MxFP4 MoE bin kernel for Adreno X1 devices (#30691)
  
  shawngu-quic
  and
  lhez
  authored
  2f2763d
- hexagon: vectorize softplus with HVX (#30435)
  huangwei858
  authored
  27c7a76
- ci: update device matrix for Snapdragon jobs by removing SM8750 (#30620)
  zhiyuan8
  authored
  6420cf5
Commits on Oct 17, 2026
- skills: add slice-model (#30677)
  ngxson
  authored
  ea0c4c2
- sycl: fix prompt-processing regression after Q4_0/Q8_0 weight reorder (#30374)
  marcusds
  authored
  f74e2fa
- model : Granite switch decision 256 options (#30693)
  gabe-l-hart
  authored
  1268eda
- ci : added ABI reports as github artifact (#30619)
  sandrohanea
  authored
  83f9389
- model : add Kolibri-1 (Aleph-Alpha) support (#30673)
  
  riteme
  and
  CISC
  authored
  62318cd
- cuda : add grouped experts top-k fusion (#29181)
  CISC
  authored
  f2f667b
- cuda : fix -Wtautological-compare in fattn vec dispatch (#30519)
  angt
  authored
  9d1e036
- tests: give the random model generator a per-tensor filler table (#30674)
  ServeurpersoCom
  authored
  10b8b75
- CUDA: faster MMVQ for low-bit types (#30517)
  
  am17an
  and
  thelittlefireman
  authored
  2744e48
- ui : Discover & Download Models from HF Hub (#29584)
  
  authored
  c0fd786
- ggml-cuda: HIP: use __builtin_amdgcn_perm for q1_0 unpacking (#29927)
  thelittlefireman
  authored
  9d1878a
- hexagon: add support for Q1_0 and Q2_0 (#30122)
  
  tboinovski1
  and
  max-krasnyansky
  authored
  797de63
Commits on Oct 18, 2026
Commits on Oct 19, 2026
Commits on Oct 20, 2026
Commits on Oct 21, 2026
- opencl: fix rope program build error with newer Clang (#30893)
  bashbaug
  authored
  339f324
- Sycl iq3 crash fix (#30604)
  anantshri
  authored
  493253f
- llama : scale moe down projection per expert token (#30861)
  ggerganov
  authored
  8baabd5
- ggml-cpu : compute f32 gelu_quick and geglu_quick without the fp16 table (#30737)
  boshjerns
  authored
  479a527
- sycl: collapse duplicate need_check launch bodies in mmq (#30739)
  Titaniumtown
  authored
  96caa80
- sycl : fix fa crash for Query head dimension mismatch (neq0 != HSK / D=2048/4096) (#30286)
  arthw
  authored
  74b115c
- opencl: add dp4a GEMM variants to vector load activation (#30753)
  lhez
  authored
  b531e4c
- hexagon: neon-vectorized and threaded repack (tiling) (#30901)
  
  max-krasnyansky
  and
  hubezhou
  authored
  0efe8e4
- metal : fix memory leaks and cleanup logic in GPU capture path (#30890)
  nikwen
  authored
  ea9e344
- llama : place d2t by its set_rows use so it can stay on the GPU (#30912)
  am17an
  authored
  eb182d8
- llama : optimize vocabulary loading (#30679)
  
  mykhailen
  and
  ggerganov
  authored
  824eaec
- vendor: update BoringSSL to 0.20261020.0 (#30889)
  cabelo
  authored
  f849095
- server : forward llama.cpp sampling params on /v1/messages (#30856)
  Dgarc359
  authored
  261700b
- mtmd: add state object for mtmd_gen (#30902)
  ngxson
  authored
  0a1ef30
- ggml-opencl : add IM2COL_3D support (#30895)
  
  kurquhar
  and
  lhez
  authored
  9b56ccf
- jinja: support hexadecimal string escapes (#30929)
  alexey-rsi
  authored
  9febe63
- sycl: q8 fast dequant (#30775)
  92es
  authored
  0756e83
- llama: create new fused op for QSA mask (#30640)
  
  am17an
  and
  ServeurpersoCom
  authored
  d454a20
- model: Add support for SenseNova U1 model (#28919)
  
  Maphist0
  and
  CISC
  authored
  7775ffb
- jinja : fix namespace attributes shadowed by dict methods (#30941)
  tdakhran
  authored
  99fb997
- models : add Flint model (FlintForCausalLM) (#30928)
  
  alexey-rsi
  and
  CISC
  authored
  a37cc46
Commits on Oct 22, 2026
Commits on Oct 23, 2026
- sycl: Add graph record and replay (#28725)
  malsbat
  authored
  7377bad
- sycl: Replace host waits in fattn-mkl with explicit dependencies (#30684)
  malsbat
  authored
  0a9ef00
- hexagon: Op Fusion refactoring, optimizations and new fusions like `BINARY_NX` (back-to-back ADDs, etc) (#30843)
  max-krasnyansky
  authored
  643b0cd
- vocab : fix Nemotron-3.5-Embed tokenization (#30974)
  
  authored
  f9599d0
- Extend probabilistic sampling for DFlash2 drafts (#30663)
  
  praneshgo
  and
  Pranesh Gonegandla
  authored
  1ab09d6
- sycl : pass device into mul_mat_uses_onednn (#31014)
  Titaniumtown
  authored
  f64a498
- CANN: missing `const` qualifiers added for consistency (#30933)
  rauletorresc
  authored
  f5027fb
- ci : add behavior check workflow (#31036)
  CISC
  authored
  f00c908
- ggml : fix time overflow in ggml_time_us (#30969)
  frozenblade1224
  authored
  4231b23
- ui : bump packages to clear npm audit (#31038)
  allozaur
  authored
  e1e3b2c
- hexagon: add CEIL op (#31006)
  max-krasnyansky
  authored
  36e36e6
- CUDA: chunk graphs to amortize graph launch (#30804)
  am17an
  authored
  280141e
- vendor : update LibreSSL to 4.4.0 [no ci] (#31049)
  angt
  authored
  da2ee1e
- qwen4exp : build the QSA selection without the kq_mask (#30965)
  sanmai
  authored
  5c31142
- allows the usage of 4-10 gpus for stepfun and laguna in tensor parallel (#24554)
  krampenschiesser
  authored
  06230cb
- CUDA: fix build after chunked CUDA graphs (#30804) (#31053)
  
  sreeram-11
  and
  JohannesGaessler
  authored
  9ff0d7e
- ggm-cuda: Enforce DP4A MMQ tile height multiple of warp size (#30837)
  thelittlefireman
  authored
  1a24df1
- hip: fix bf16 WMMA operand vector type for gfx11/gfx12 (#30885)
  waltercool
  authored
  9145ddb
- common : allow concurrent downloads across processes (#29312)
  angt
  authored
  21ec307
- hexagon: add SGN op (#30346)
  
  kurquhar
  and
  max-krasnyansky
  authored
  0ae0e19
- mtmd : add Soprano TTS support (#28607)
  hans00
  authored
  9ae73e6
Commits on Oct 24, 2026
Commits on Oct 25, 2026
- model: DSV4.1 flash (#31086)
  
  am17an
  and
  danielhanchen
  authored
  74bca39
- chat : reject JSON nested more than 128 levels before walking it (#30242)
  
  ykhrustalev
  and
  ngxson
  authored
  b3fa699
- ci : update CPU job assignments (#31136)
  ggerganov
  authored
  344be0a
- opencl: add initial q2_k support (#30999)
  
  lhez
  and
  wanghqc
  authored
  b083c38
- opencl: add topk and lightning indexer (#31026)
  lhez
  authored
  9a4580b
- hexagon: generic dims collapse/reshape for unary, binary and matmul ops (#31071)
  max-krasnyansky
  authored
  f0fb598
- vulkan : one warp tile per subgroup in the small matmul tiles (#30106)
  ianloic
  authored
  459a1ba
- vulkan: Add a larger tile size for conv3d (coopmat2 only) (#29920)
  jeffbolznv
  authored
  e321bda
- ggml-cuda: fuse qwen MRoPE + set-rows (#31025)
  am17an
  authored
  a55c16e
- CUDA: fuse FFN gate/up matmuls and GLU in MMQ with interleaved gate/up rows (#29948)
  anujj
  authored
  c49edf0
- tools : reject truncated GGUF tensor data in merge (#31044)
  
  shreya-11733
  and
  Shreya Agrawal
  authored
  7790a92
- llama : respect no_alloc in all memory caches (#31087)
  ggerganov
  authored
  fa3645f
- grammar: limit recursion depth to prevent stack overflow (#29495) (#29755)
  Hrushikesh-ramilla
  authored
  f1168ac
- llama-quant : fix 4D tensor quantization using ggml_nrows() and name-based expert detection (#30820)
  anthony-wss
  authored
  5c965fd
- llama: fix exhausting CPU and MEM caused by dry sequence breaker (#31035)
  trunk-trick
  authored
  63412d8
- ci : disable test-llama-context with fa off in the webgpu (#30977)
  yomaytk
  authored
  1a0b56c
- qwen: merge tensors on convert (#31028)
  am17an
  authored
  2280f0f
- tests : adjust default threads for test-llama-context (#31157)
  ggerganov
  authored
  6050b46
- common, grammar: add recursive depth limit, make it a requirement (#31153)
  
  authored
  48c9f08
- cuda : use std::string instead of VLA for archName (#30463)
  angt
  authored
  d854d98
Commits on Oct 29, 2026

Commits

Commits on Oct 15, 2026

Commits on Oct 16, 2026

Commits on Oct 17, 2026

Commits on Oct 18, 2026

Commits on Oct 19, 2026

Commits on Oct 20, 2026

Commits on Oct 21, 2026