Repository navigation

tests: per-layer recompute under the exact walk takes the same step in a smaller chunk graph - #50

Merged

joelteply merged 1714 commits into

masterfrom

train/recompute-under-exact-walk

Oct 10, 2026

Merged

tests: per-layer recompute under the exact walk takes the same step in a smaller chunk graph#50
joelteply merged 1714 commits into
masterfrom
train/recompute-under-exact-walk

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

Commits on Aug 22, 2026
- ci : add older, min and dry-run options to ccache-clear (#27504)
  ggerganov
  authored
  3a653fe
- mtmd: support dots3-note vision+audio (#27524)
  
  ngxson
  and
  CISC
  authored
  54ee5ee
- model : support DSpark for bailingmoe3 (#27508)
  btw616
  authored
  2115b73
- ci : Restore ROCm job for Ubuntu (#27399)
  
  authored
  e85caa8
- arg: remove -no-cnv from cli [no ci] (#27542)
  ngxson
  authored
  9fee29e
- fit: also take into account n_streams (#27496)
  
  ngxson
  and
  ServeurpersoCom
  authored
  2fb989b
- common: add json.h abstraction (#27511)
  ngxson
  authored
  d9f918d
- mtmd: use ggml_rope_set_offset (#27521)
  ngxson
  authored
  b21e4de
- vulkan : added the PAD_REFLECT_1D operation (#26586)
  
  safiullah3915
  and
  jeffbolznv
  authored
  3f545be
- common: json.h: fix clang lto (#27575)
  ngxson
  authored
  70adb1b
Commits on Aug 23, 2026
- cuda : add POOL_1D support (#27573)
  amankarki151
  authored
  29ea941
- vendor : update subprocess.h (#27409)
  cabelo
  authored
  6657ded
- ui: Chat Conversation Tabbed navigation (#27263)
  
  allozaur
  and
  ServeurpersoCom
  authored
  8144f31
- [Tensor parallel] Fix meta tensor split state propagation (#27574)
  gaugarg-nv
  authored
  d337192
- DeepseekV4: fix rollback with multi-seq (#26756)
  
  am17an
  and
  ggerganov
  authored
  b0539c4
- common : skip device_info loop if it's not going to be printed (#26692)
  wolfpld
  authored
  ba8e0ed
- server : add LLAMA_SERVER_SLOTS_N_DIFF (#27600)
  ggerganov
  authored
  e8eed45
- contrib : recommend waiting for CI before merging (#27603)
  nikwen
  authored
  a278dce
- ci : add test-llama-archs tensor split for Metal (#27598)
  ggerganov
  authored
  95b8e33
- mtmd: use pillow-accurate algo, correct resize_algo for all models (#27594)
  ngxson
  authored
  56db501
- test: move tools/parser to tests (#27548)
  ngxson
  authored
  4a08fa2
- test : fix multi-GPU server tests (#27614)
  ggerganov
  authored
  8d9af25
- fix: Change chat tabs nav shortcuts (#27609)
  allozaur
  authored
  d05f895
- readme : update links (#27617)
  ggerganov
  authored
  ccc8fd2
- model : support MTP in GLM-4.5-Air (#26534)
  jacekpoplawski
  authored
  c060ca9
Commits on Aug 24, 2026
Commits on Aug 25, 2026
- webgpu : fix handling of infinity values during ARGSORT and TOP_K (#27538)
  
  fairydreaming
  and
  sszymczy
  authored
  5ea87dd
- sycl : mark tq2_0 as not supported (#27660)
  arthw
  authored
  814d84b
- grammar : parse \- in char classes as literal hyphen (#27591)
  NIXKnight
  authored
  eb25b72
- ggml : bump version to 0.22.0 (ggml/1607)
  ggerganov
  committed
  9388236
- sync : ggml
  ggerganov
  committed
  81191af
- llama.cpp : bump version to 0.3.0 (#27696)
  ggerganov
  authored
  c1d0e7a
- metal : null-check buffer alloc to fix OOM crash (#25371)
  
  ykhrustalev
  and
  ggerganov
  authored
  3737e41
- ui: ESLint config updates (#27700)
  allozaur
  authored
  f1357e4
- ci : store ccache on HF buckets (test with cuda-ubuntu for now) (#27699)
  CISC
  authored
  790b571
- chat : scope qwen3-coder workarounds (#27679)
  aldehir
  authored
  0cc5b14
- server : reject prefilled assistant messages with tool calls (#27626)
  kyo-zzz
  authored
  1729ed5
- devops: use GGML_NATIVE=OFF for OpenVINO (#27338)
  
  krzsztf
  and
  CISC
  authored
  0a5ac49
- rpc: support apple RDMA as an RPC transport (#26421)
  
  ryan5rdx
  and
  Ryan Churaman
  authored
  b114b47
- ci : update OS used for ROCM to Ubuntu 24.04 (#27681)
  superm1
  authored
  eab8ee4
- kleidiai: Rework KleidiAI Build System/Integration (#26077)
  JonathanC-ARM
  authored
  d222767
Commits on Aug 26, 2026
- ggml-meta: propagate buffer usage and call init on the new tensors (#27586)
  max-krasnyansky
  authored
  5d5cb4c
- ggml-metal: add chunked SSD MMA for Mamba-2 prefill optimization (#26647)
  
  authored
  11cd988
- conversion : fix Nemotron 3.5 Lightning layers (#27729)
  danbev
  authored
  dac869b
- ci : make cache bucket public (#27728)
  CISC
  authored
  da9b5d6
- cuda: unblock mmq for MoE on sm_60 (#26264)
  dfriehs
  authored
  fc35562
- ci: Clean up UI builds from releases (#27706)
  allozaur
  authored
  4d19b28
- gguf: add qwen4exp (Qwen3.8-Flash-Next) arch and converter
  danielhanchen
  committed
  6e5b8b7
- llama: load qwen4exp (Qwen3.8-Flash-Next) hparams and tensors
  danielhanchen
  committed
  331fa7e
- qwen4exp: shorten comments
  danielhanchen
  committed
  058b7a1
- llama: qwen4exp text graph with hyper-connections, GDN and MoE
  danielhanchen
  committed
  c9ad86f
- llama: qwen4exp PLE n-gram hash embedding
  danielhanchen
  committed
  ad4fa3f
- llama: carry the qwen4exp PLE conv state across ubatches
  danielhanchen
  committed
  c89e67b
- llama: fix the qwen4exp PLE conv state and unblock test-llama-archs
  danielhanchen
  committed
  ddf0980
- llama: optional indexer key cache in llama_memory_hybrid
  danielhanchen
  committed
  ff24f38
- llama: QSA sparse attention for qwen4exp
  danielhanchen
  committed
  c88c916
- llama: give the qwen4exp indexer cache the attention cache's slots
  danielhanchen
  committed
  1b59910
- tests: record what the qwen4exp arch-test skip actually observes
  danielhanchen
  committed
  dd89627
- tests: cover qwen4exp in test-llama-archs
  danielhanchen
  committed
  40e9041
- convert: stream the qwen4exp PLE table instead of concatenating it
  danielhanchen
  committed
  53c0f62
- quantize: fall back to F16 for 32-block types with an odd ncols
  danielhanchen
  committed
  570c26d
- quantize: let --tensor-type name per_layer_token_embd
  danielhanchen
  committed
  4597a86
- quantize: size the output buffer exactly instead of nelements * 4
  danielhanchen
  committed
  25aa77a
- qwen4exp: hash the image placeholder for multimodal batches
  danielhanchen
  committed
  eb95d12
- qwen4exp: support a non-unified KV cache in QSA
  
  danielhanchen
  authored and
  danielhanchen
  committed
  bea3b12
- llama: keep the qwen4exp top-k attention mask arch-local
  danielhanchen
  committed
  7fee670
- llama: hold the qwen4exp indexer cache in a new llama_memory_hybrid_idx
  danielhanchen
  committed
  035e227
- Merge upstream b10636 into CambrianTech fork (base was b10143, 493 upstream commits)
  joelteply
  committed
  2ae88c6
- restore per-layer head-geometry APIs lost in the b10636 merge
  joelteply
  committed
  67b2868
- fix: drop duplicated kimi-k3 hparams block from the b10636 merge
  
  joelteply
  and
  claude
  committed
  986754c
Commits on Aug 27, 2026
- llama: save and restore the qwen4exp indexer KV cache
  danielhanchen
  committed
  cfbdc0a
- llama: make the qwen4exp PLE n-gram history per context and serialise it
  danielhanchen
  committed
  d22d2be
- qwen4exp: tidy comments and simplify image token read
  
  danielhanchen
  authored and
  danielhanchen
  committed
  d4a943f
- qwen4exp: support a quantized KV cache in the QSA attention path
  
  danielhanchen
  authored and
  danielhanchen
  committed
  0ac4b18
- llama: give qwen4exp a large-graph node budget
  
  danielhanchen
  authored and
  danielhanchen
  committed
  c52ed2a
- qwen4exp: drop an unused variable that breaks -Werror builds
  
  danielhanchen
  authored and
  danielhanchen
  committed
  6a69a0c
- quantize: dequantize and quantize large tensors in row bands
  
  danielhanchen
  authored and
  danielhanchen
  committed
  ef9fa1b
- llama: segment the qwen4exp fused QKV for tensor split
  danielhanchen
  committed
  5674c73
- llama: fix the qwen4exp PLE history seq_rm(-1) iterator invalidation and the fatal-warning build
  danielhanchen
  committed
  24ea62d
- qwen4exp: include llama-impl.h explicitly for llama_mul_mat_hadamard
  
  danielhanchen
  authored and
  danielhanchen
  committed
  0b19188
- convert: fix the qwen4exp lint and type-check failures
  danielhanchen
  committed
  213df58
- llama: give the qwen4exp indexer cache its own tensor names
  
  danielhanchen
  authored and
  danielhanchen
  committed
  fbe1773
- qwen4exp: double the Q split granularity for tensor parallelism
  
  danielhanchen
  authored and
  danielhanchen
  committed
  6c5afc8
- Merge llama.cpp PR #27742 (qwen4exp: Qwen3.8-Flash-Next — GDN+QSA hybrid, n-gram embeddings, 512-expert MoE) into CambrianTech fork
  joelteply
  committed
  52f761b
Commits on Aug 28, 2026
- Merge k3-adopt (moe-gather #23 Cut-2) into merge/qwen4exp — the union pin
  joelteply
  committed
  ae438cd
- fix(ggml-backend): include <memory> for the MoeGatherTables unique_ptr pool
  joelteply
  committed
  920eef0
Commits on Sep 2, 2026
- server: don't defer SLOT_SAVE on a live slot (concurrent warm-KV save)
  
  joelteply
  and
  claude
  committed
  878db97
Commits on Sep 3, 2026
- fix(server): bind slot page-ops to their slot so a deferred RESTORE is serviced when ITS slot frees (the 27/27 restore-fail)
  
  joelteply
  and
  claude
  committed
  0a637ba
Commits on Sep 4, 2026
- continuum: llama_model_get_tensor — public mutable weight accessor for K3 per-expert paging
  
  joelteply
  and
  claude
  committed
  c08a004
Commits on Sep 5, 2026
Commits on Sep 13, 2026
Commits on Sep 26, 2026
Commits on Sep 27, 2026
Commits on Sep 28, 2026
Commits on Oct 5, 2026
Commits on Oct 6, 2026
Commits on Oct 7, 2026
Commits on Oct 10, 2026

Commits

Commits on Aug 22, 2026

Commits on Aug 23, 2026

Commits on Aug 24, 2026

Commits on Aug 25, 2026

Commits on Aug 26, 2026

Commits on Aug 27, 2026

Commits on Aug 28, 2026

Commits on Sep 2, 2026

Commits on Sep 3, 2026

Commits on Sep 4, 2026

Commits on Sep 5, 2026

Commits on Sep 13, 2026

Commits on Sep 26, 2026

Commits on Sep 27, 2026