llama.cpp

millaguie/llama.cpp

Fork 0

9303cdd8d3 security : clarify about AI-generated reports (#26579) Xuan-Son Nguyen 2026-08-05 13:27:06 +02:00
a035a88878 server: Adding spec-decode counters to /metrics endpoint (#26389) Bhavik Sharda 2026-08-05 16:06:01 +05:30
020760adfc convert: Add endianness conversion for Q1 and TQ2 quantizations (#26618) Andreas KrebbelandSigbjørn Skjæret 2026-08-05 12:06:09 +02:00
61881b1f7f vendor : apply patches for subprocess.h (#26606) Xuan-Son Nguyen 2026-08-05 11:26:20 +02:00
3e3a7a416d ui: show generation statistics by default in chat settings (#26624) Aleksander Grygier 2026-08-05 11:03:23 +02:00
d52ec04a66 build : remove GGML_METAL_USE_BF16 from all build scripts (#26604) Niklas Wenzel 2026-08-05 10:44:34 +02:00
e031d95679 ui: Update vulnerable packages + cleanup Storybook config (#26607) Aleksander Grygier 2026-08-05 08:06:37 +02:00
6ea215d171 Prefer npm ci over install for security (#26601) Evan Huus 2026-08-04 18:14:22 -04:00
4308a4f035 server: decode Windows OEM output to UTF-8 in built-in tools (#26597) Pascal 2026-08-04 22:24:55 +02:00
474c92e722 mtmd: correcting duplicate empty audio chunks for short inputs (#26536) Abhinay Krishna 2026-08-04 16:05:56 -04:00
a6aa6f5450 sampler : remove "full-context windows" from history-based samplers (#26524) Oliver Simons 2026-08-04 20:28:55 +02:00
76c956c137 gguf-split: Add option to delete split parts during merge (#26538) Guilherme QuintinoandSigbjørn Skjæret 2026-08-04 19:27:47 +01:00
2f56fc3431 ui: CWD for agent (#26518) Aleksander GrygierandPascal 2026-08-04 19:05:48 +02:00
0713275082 mtmd: support Qwen3-TTS (note: breaking change to llama-tts binary) (#26254) Xuan-Son NguyenandPascal 2026-08-04 17:26:15 +02:00
1c3c9674de models : fix dflash wo_a reshape on load (#26577) Georgi Gerganov 2026-08-04 16:56:49 +03:00
6b5224cfcc ci: fix pre-built binaries no longer working on macOS 15 and below (#26375) Niklas Wenzel 2026-08-04 15:03:38 +02:00
7bd8282c37 speculative : refactor enabled configs common_speculative_init (#26510) Daniel Bevenius 2026-08-04 13:17:15 +02:00
5788b510a1 gguf-py: validate n_dims and guard against uint64 overflow in reader (#25401) hcl 2026-08-04 17:12:48 +08:00
2e17f69ef4 sync : ggml Georgi Gerganov 2026-08-04 11:54:13 +03:00
15831f579a ggml : bump version to 0.18.1 (ggml/1578) Georgi Gerganov 2026-08-04 11:44:52 +03:00
b5746d28ce convert : add missing return after setting tekken vocab (#25947) Angel Galindo 2026-08-04 01:41:18 -07:00
f26efa02a7 vulkan backend ops: implemented GATED_LINEAR_ATTN (#25601) Pranav Uttarkar 2026-08-04 03:40:54 -05:00
cf06ad7dfe vocab : validate plamo2 byte tokens (#26511) Sigbjørn Skjæret 2026-08-04 10:40:02 +02:00
b06fbc968b convert : import bytes_to_unicode from convert_slow_tokenizer (#26217) Caleb DeLeeuw 2026-08-04 00:34:30 -07:00
1269cb1ff1 model : allow reshape of tensors during load (#26531) Georgi Gerganov 2026-08-04 09:06:44 +03:00
935cad6497 llama : move n_vocab from llama_sampler_data to penalty_sampler (#26520) Oliver Simons 2026-08-04 08:02:49 +02:00
22dc605c4e ci: fix vulkan llvmpipe runs (#26533) Eve 2026-08-04 03:28:57 +00:00
6c8dcaa7ae sycl: parallelize the non-contiguous concat kernel (#25852) Titaniumtown 2026-08-03 19:08:05 -07:00
66fa168a56 Extended SYCL oneDNN SDPA to non-FP16 KV caches (Q4_0–Q8_0 and FP32) (#25874) Ozymandias_EBONandClaude 2026-08-03 21:07:23 -05:00
0ef6e55edb chat : add new template for DeepSeek V4 Flash 0731 (#26398) Thiago Padilha 2026-08-03 19:59:11 -03:00
94bc47f280 vendor : update cpp-httplib to 0.52.0 (#26485) Alessandro de Oliveira Faria (A.K.A.CABELO) 2026-08-03 19:30:42 -03:00
fe2adf0e72 vendor : update BoringSSL to 0.20260803.0 (#26523) Alessandro de Oliveira Faria (A.K.A.CABELO) 2026-08-03 15:31:15 -03:00
57c092139a model : support MTP in GLM-4.7-Flash (#24868) jacekpoplawski 2026-08-03 20:27:52 +02:00
ee0445c99c tests: add model resolution test on synthetic repo listings (#26172) PascalandXuan Son Nguyen 2026-08-03 18:58:15 +02:00
99111b19ce server: add get_info tool (#26522) Xuan-Son NguyenandPascal 2026-08-03 18:51:02 +02:00
e8e06f78e2 vocab : validate default special token ids (#26506) Sigbjørn Skjæret 2026-08-03 17:40:53 +02:00
dbadb68eec ggml: use dynamic allocation for split graph inputs (#22789) AgoraPeteandGeorgi Gerganov 2026-08-03 17:03:14 +02:00
39eab74a05 opencl: route large q6_K lm_head to the flat GEMV (#26427) Hongqiang Wang 2026-08-03 07:36:19 -07:00
c50b34a1e0 graph : fix unused input tensors in minimax m3 graph (#26519) Georgi Gerganov 2026-08-03 17:32:01 +03:00
67d5978bb1 model: M3: Move MSA into a new memory implementation (#26338) timkhronosandGeorgi Gerganov 2026-08-03 15:30:08 +02:00
563dec81c1 llama : allocate indexer cache only in "full" indexer layers (#26474) fairydreamingandStanisław Szymczyk 2026-08-03 14:56:30 +02:00
96278e39fc CUDA: Add backend sampler for penalties sampler (#25262) Konrad Moren 2026-08-03 14:26:09 +02:00
9bd4c09ea5 CUDA: Fix data-races when reusing SMEM in block_reduce (#26385) Oliver Simons 2026-08-03 14:22:44 +02:00
0b14b87d7c server: add notice for upcoming default port change 8080 --> 9931 (#26508) Xuan-Son Nguyen 2026-08-03 12:45:24 +02:00
f2b52a87e8 server: (tools) add x-tool-cwd header (#26420) Xuan-Son Nguyen 2026-08-03 10:47:21 +02:00
4ed2b13f75 model: MTP support for Qwen3-Next (#25589) Masashi YoshimuraandSigbjørn Skjæret 2026-08-03 17:15:01 +09:00
2b63e0610b llama : MTP support for DeepSeek V3.2 (#26457) fairydreamingandStanisław Szymczyk 2026-08-03 08:25:01 +02:00
1464c62d88 metal: implement DSv4 Lightning Indexer (#25893)
Thiago Padilha forforever73 Georgi Gerganov
2026-08-03 01:33:37 -03:00
221f0f6356 metal : add SILU_BACK (#25982) Talha Adnan 2026-08-02 14:39:28 -05:00
9d21b57f2e metal : add F16 support for bin ops (#26465) Georgi Gerganov 2026-08-02 22:28:17 +03:00
0ab9d6fed7 opencl: limit local workgroup size for GLU operation (#26383) mgroeber9110 2026-08-02 20:44:00 +02:00
fffbcbdb9d metal: implement DeepSeek V4 hyper-connections (#26459) Georgi GerganovandThiago Padilha 2026-08-02 21:06:02 +03:00
bb4e0e1b3f common: support the DSpark sidecar resolution (#26458) Pascal 2026-08-02 19:25:27 +02:00
3581ba0cf5 convert: add option to create separate dspark GGUF (#26452) Aman Gupta 2026-08-02 23:16:31 +08:00
c745be2a2c opencl: bugfix increment ref_count in ggml_backend_opencl_init() (#26162) akleine 2026-08-02 15:43:00 +02:00
596a5795bd DeepseekV4 MTP + DSpark (#25784) Aman Gupta 2026-08-02 20:55:34 +08:00
f5919bf458 chat : add qwen3 specialized parser (#26252) Aldehir RojasandBart de Boer 2026-08-02 04:13:20 -05:00
272700b360 sycl: fix classification of iGPUs (#26105) KyleHagy 2026-08-02 00:10:32 -07:00
75587a05b3 model : load MiMo V2 MTP tensors only if used (#26412) Sigbjørn Skjæret 2026-08-02 09:03:05 +02:00
7a2db1a0cf ggml-webgpu: add support for f16 repeat (#26307) Masashi Yoshimura 2026-08-02 15:28:31 +09:00
11924d4c17 test: fix some CI errors (#26415) Xuan-Son Nguyen 2026-08-02 00:16:29 +02:00
a7a6d0d269 vulkan: extend topk_moe fusion to support sqrt(softplus) (#26124) Jeff Bolz 2026-08-01 14:18:07 -05:00
815a2a5915 vendor : update BoringSSL to 0.20260730.0 (#26353) Alessandro de Oliveira Faria (A.K.A.CABELO) 2026-08-01 15:53:00 -03:00
89482bd665 agents: clarify comment style and jinja knowledge (#26405) Xuan-Son Nguyen 2026-08-01 18:45:46 +02:00
c629da565c cli : persist reasoning_content in chat history (#26362) Nico 2026-08-01 12:03:32 -04:00
de699957b9 mtmd: add minicpmv46 downsample (#25993) tc-mb 2026-08-01 19:38:36 +08:00
ddd4ec1428 chat : enable tool call in thinking for DS4 (#26269) Piotr Wilkin (ilintar) 2026-08-01 07:13:07 +02:00
876a432116 vulkan: add POOL_1D op (#25431) Anand Patil 2026-07-31 09:48:58 -05:00
eb41d503ba vulkan: Introduce driver version check for Windows Intel GPU to mitigate crashing (#25192) Masato Nakasaka 2026-07-31 23:26:37 +09:00
db7d8b24b5 mtmd: add n_embd_head (#26342) Xuan-Son NguyenandDaniel Han 2026-07-31 15:30:19 +02:00
a09d8abf8c Support rotated kv cache quant (#26180) timkhronos 2026-07-31 15:06:40 +02:00
82dbc4f017 llama : load MTP tensors only if they are really used (#26296) fairydreamingandStanisław Szymczyk 2026-07-31 14:57:02 +02:00
6f3c0a790b vulkan: update vulkan sdk to 1.4.357.0 (#26303) Jeff Bolz 2026-07-31 13:27:03 +01:00
000547513f server: correct accepted tokens when need draft token replay (#26320) Ruixiang WangandGeorgi Gerganov 2026-07-31 10:16:17 +02:00
15e755f30d cuda: extract Q2_0 elements via __byte_perm (#25603) David Friehs 2026-07-31 10:15:44 +02:00
9d9a6d29f6 SYCL: add oneMKL GEMM flash attention for XMX-accelerated prompt proc… (#25025)
Ozymandias_EBON Neo Zhang Claude Code using DeepSeek-V4-Pro
2026-07-31 02:43:16 -05:00
d5d3e05bf8 [SYCL] support the missed types in cpy (#26005) Neo Zhang 2026-07-31 15:25:16 +08:00
69e62fc77c llama : enforce the same K and V cache types for DeepSeek V4; enable FA if V cache is quantized (#25871) fairydreamingandStanisław Szymczyk 2026-07-31 09:03:30 +02:00
1e22599522 ggml-zendnn : group matmul direct API for mul_mat_id (#25918) Sachin Sharma 2026-07-31 12:10:52 +05:30
1c5b89ff63 sycl : support dev2dev memcpy by DEV2DEV_MEMCPY_FORWARD (#26234) Neo ZhangandNeo Zhang Jianyu 2026-07-31 14:20:28 +08:00
a2be61dc87 [SYCL] Support q2 mul_mat (#26231) Neo Zhang 2026-07-31 14:19:41 +08:00
1553725965 sycl: fuse RMS_NORM + MUL (#26015) Titaniumtown 2026-07-30 23:17:53 -07:00
8f4646a63e ggml-webgpu: improve flash_attn_vec for quantized KV at long contexts (#25956) Masashi Yoshimura 2026-07-31 15:08:40 +09:00
5f55650a78 mtmd: add lanczos resize method [no release] (#26341) Xuan-Son Nguyen 2026-07-30 21:59:49 +02:00
b4ca032ae3 server: support inp embd to generate next token (#26313) Xuan-Son Nguyen 2026-07-30 21:40:38 +02:00
ea63b4d32e vulkan: Support quantized concat (#25684) Jeff Bolz 2026-07-30 19:11:32 +01:00
958d9c0b61 Test support for alternative conv layout (#25617) pmaybank 2026-07-30 18:14:16 +01:00
432d7ffe2c llama-context : sync pending async copies before clearing embd_seq (#25676) o7si 2026-07-31 00:48:00 +08:00
47f686f53f tests : avoid building get-model.cpp many times (#26317) Georgi Gerganov 2026-07-30 19:34:04 +03:00
e1a1abb787 ggml-cuda: Allow transpose-free gemmv computation (#26171) Robert Esclapez 2026-07-30 15:39:46 +02:00
6b36c23056 readme : refresh (#26280) Georgi Gerganov 2026-07-30 16:14:37 +03:00
9ebfc3a8cf sync : ggml Georgi Gerganov 2026-07-30 15:43:44 +03:00
6a4c3357c8 ggml : bump version to 0.18.0 (ggml/1576) Georgi Gerganov 2026-07-30 15:42:47 +03:00
9b2a088819 CUDA: add Q2_0 support (#25707) Pasha Khosravi 2026-07-30 02:33:25 -07:00
b2f221684f Remove custom cpu op from the M3 graph, express with stock ops (#26297) timkhronos 2026-07-30 10:30:18 +02:00
d0bfb19812 metal: fix memory unwire if model is freed without any GPU operations (#26082) Niklas WenzelandYiChen Lv 2026-07-30 10:11:27 +02:00
21a5f5b7f9 ui: IndexedDB and Conversations data fixes (#26278) Aleksander Grygier 2026-07-30 10:10:37 +02:00
32703b42d6 ggml : Fix issue with kleidiai ci and stringop overflow warning (#26277) Jonathan Clohessy 2026-07-30 07:17:30 +01:00
a6a77bc48d [UT] enhance UT to show all real unsupported backends (#25234) Neo ZhangandGeorgi Gerganov 2026-07-30 14:04:58 +08:00
64d528be72 mimo2: address MTP review feedback (#26228) Tunahanandtnhnyc 2026-07-30 05:55:58 +02:00

1 2 3 4 5 ...

Commit Graph Select branches Hide Pull Requests dt3 master Mono Color

Commit Graph

Select branches

Hide Pull Requests

dt3

master