upstream-archive/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-07-28 21:50:56 -05:00
Code Issues Packages Projects Releases Wiki Activity

Commit Graph

  • 2aa777d86d examples : switch retrieval to llama_encode (#13685) b5446 Sigbjørn Skjæret 2025-05-21 16:57:38 +02:00
  • eb0f5c28d3 gguf-py : display the invalid gguf type (#13687) Emmanuel Ferdman 2025-05-21 17:33:54 +03:00
  • cf4cb59e64 ggml : add ggml_gelu_erf() (#13667) b5444 Xuan-Son Nguyen 2025-05-21 16:26:33 +02:00
  • 0d5c742161 server : Add the endpoints /api/tags and /api/chat (#13659) b5443 Robin Davidsson 2025-05-21 15:15:27 +02:00
  • 42158ae2e8 server : fix first message identification (#13634) b5442 Dorin-Andrei Geman 2025-05-21 16:07:57 +03:00
  • 797f2ac062 kv-cache : simplify the interface (#13660) b5441 Georgi Gerganov 2025-05-21 15:11:13 +03:00
  • b44890df2e model : disable SWA for Phi models (#13676) b5440 Georgi Gerganov 2025-05-21 13:09:21 +03:00
  • 33983057d0 musa: Upgrade MUSA SDK version to rc4.0.1 and use mudnn::Unary::IDENTITY op to accelerate D2D memory copy (#13647) b5439 R0CKSTAR 2025-05-21 09:58:49 +08:00
  • fb1cab201c vulkan: fix warnings (#13626) b5438 Eve 2025-05-20 21:35:16 +00:00
  • b7a17463ec mtmd-helper : bug fix to token batching in mtmd (#13650) b5437 l3utterfly 2025-05-21 00:55:30 +08:00
  • be0239693c model : fix llama4 graph (#13663) b5436 Georgi Gerganov 2025-05-20 19:21:04 +03:00
  • a4090d1174 llama : remove llama_kv_cache_view API + remove deprecated (#13653) b5435 Georgi Gerganov 2025-05-20 16:13:16 +03:00
  • b69f1647f9 CUDA: skip fully masked-out KV in FA vec kernel (#13584) b5434 Johannes Gäßler 2025-05-20 14:45:07 +02:00
  • 759e37b0d8 tests : avoid github urls due to throttling (#13654) Sigbjørn Skjæret 2025-05-20 12:03:17 +02:00
  • 4245e622e0 sycl: disable reorder for sycl mulmat (#13536) b5432 Svetlozar Georgiev 2025-05-20 10:34:15 +01:00
  • c9c64dee57 Set GLM4 blk.*.attn_output.weight, kqv_out-* matmul to GGML_PREC_F32 to fix infinity values in output (#13639) b5431 0cc4m 2025-05-20 10:11:56 +02:00
  • c00a2634be metal : fix typo in FA kernel comments (#13651) b5430 Georgi Gerganov 2025-05-20 10:41:40 +03:00
  • e298d2fbd0 kv-cache : add SWA support (#13194) b5429 Georgi Gerganov 2025-05-20 08:05:46 +03:00
  • f0adb80bf7 CANN: Update CANN model support (#13162) Xinpeng Dou 2025-05-20 11:43:43 +08:00
  • f7c9429c85 sycl : Overcoming workaround for mmap() allocation on Windows (#13482) b5427 Nicolò Scipione 2025-05-20 02:54:43 +02:00
  • 1dfbf2cf3a common : add load_progress_callback (#13617) b5426 psocolovsky 2025-05-19 21:17:36 +02:00
  • b06a954bbc llama_encode : only force non-causal attention for enc-dec models jared/permit-causal-encode Jared Van Bortel 2025-05-19 13:38:36 -04:00
  • 8960efd0a6 Vulkan: Add f32 accumulator support to quantized mul mat to fix GLM4 32B incoherence (#13607) b5425 0cc4m 2025-05-19 17:54:08 +02:00
  • 725f23f1f3 sycl : backend documentation review (#13544) Alberto Cabrera Pérez 2025-05-19 14:38:20 +01:00
  • 92ecdcc06a mtmd : add vision support for llama 4 (#13282) b5423 Xuan-Son Nguyen 2025-05-19 13:04:14 +02:00
  • f71f40a284 ci : upgraded oneAPI version in SYCL workflows and dockerfile (#13532) b5422 Alberto Cabrera Pérez 2025-05-19 11:46:09 +01:00
  • d30cb5a7fa sync : ggml b5421 Georgi Gerganov 2025-05-19 12:50:29 +03:00
  • 6c35981a64 mnist: fix segmentation fault (ggml/1227) Johannes Gäßler 2025-05-19 09:33:35 +02:00
  • 8b5e19aea6 ggml : fix apple OS check in ggml_print_backtrace (ggml/1229) Diego Devesa 2025-05-18 18:30:13 -07:00
  • 60aea028b5 ggml : Fix missing backtrace on Linux (ggml/1228) Daniel Tang 2025-05-17 19:06:26 -04:00
  • 9c55e5c5c2 fix: check model pointer validity before use (#13631) b5417 Nick 2025-05-19 18:25:41 +08:00
  • 33d7aed4a8 CANN: Support MOE Model MUL_MAT_ID (#13042) b5416 Chenguang Li 2025-05-19 14:21:17 +08:00
  • 6a2bc8bfb7 server : added --no-prefill-assistant flag (#13608) b5415 Isaac McFadyen 2025-05-17 17:59:48 -04:00
  • e3a7cf6c5b cmake: use the current build config for vulkan-shaders-gen (#13595) b5414 Gilad S. 2025-05-17 21:26:43 +03:00
  • 518329b2d4 parallel : add option for non-shared and larger prompts (#13598) Georgi Gerganov 2025-05-17 12:58:55 +03:00
  • 2f5a4e1e09 vulkan: move common FA code to flash_attn_base.comp (#13556) b5412 Jeff Bolz 2025-05-17 16:14:55 +09:00
  • 4f41ee11d6 vulkan: use scalar FA rather than coopmat2 when N==1 (#13554) b5411 Jeff Bolz 2025-05-17 15:35:47 +09:00
  • 3e0be1cace llguidance : official v0.7.20 release (no actual changes) [noci] (#13594) b5410 Z 2025-05-16 14:56:28 -06:00
  • 6aa892ec2a server : do not return error out of context (with ctx shift disabled) (#13577) b5409 Xuan-Son Nguyen 2025-05-16 21:50:00 +02:00
  • aea9f8b4e7 webui : improve accessibility for visually impaired people (#13551) Xuan-Son Nguyen 2025-05-16 21:49:01 +02:00
  • 06c1e4abc1 readme : add list of dependencies and their license (#13591) Xuan-Son Nguyen 2025-05-16 20:04:18 +02:00
  • 415e40a357 releases : use arm version of curl for arm releases (#13592) b5406 Diego Devesa 2025-05-16 10:36:51 -07:00
  • 654a67794f metal : add FA-vec kernel for head size 64 (#13583) b5405 Georgi Gerganov 2025-05-16 20:32:58 +03:00
  • 5364ae4ba5 llama : print hint when loading a model when no backends are loaded (#13589) b5404 Diego Devesa 2025-05-16 07:38:07 -07:00
  • 7c07ac244d ci : add ppc64el to build-linux-cross (#13575) Sigbjørn Skjæret 2025-05-16 14:54:23 +02:00
  • 0a338ed013 sycl : fixed compilation warnings (#13582) b5402 Łukasz Ślusarczyk 2025-05-16 12:15:29 +02:00
  • bc098c3cf0 minja: sync (qwen3) (#13573) b5401 Olivier Chafik 2025-05-15 23:29:10 +01:00
  • c6a2c9e741 gguf : use ggml log system (#13571) b5400 Diego Devesa 2025-05-15 10:13:11 -07:00
  • 07ad2b6db3 gguf-py : fix disconnect-before-connect in editor-gui (#13569) Daniel Tang 2025-05-15 12:47:10 -04:00
  • c531edfa34 convert : fix conversion for llama 4 (#13567) Xuan-Son Nguyen 2025-05-15 17:40:07 +02:00
  • 02cdd2d8b0 sycl: simplify bin_bcast_kernel (#13383) Atharva Dubey 2025-05-15 16:39:52 +01:00
  • 64bb51cf90 sycl: reordered Q4_K MMVQ (#13109) Svetlozar Georgiev 2025-05-15 16:35:44 +01:00
  • 9c404ed54c sycl: use oneDNN for matrices multiplication (#12972) b5395 Łukasz Ślusarczyk 2025-05-15 16:53:41 +02:00
  • 6c8b91500e llama-bench : fix -ot with dl backends (#13563) b5394 Diego Devesa 2025-05-15 06:46:55 -07:00
  • 3cc1f1f1d2 webui : handle PDF input (as text or image) + convert pasted long content to file (#13562) Xuan-Son Nguyen 2025-05-15 14:24:50 +02:00
  • c753d7bed0 server : proper error handling for missing elements in messages array (OpenAI compatible backend) (#13540) b5392 Piotr Wilkin (ilintar) 2025-05-15 08:40:58 +02:00
  • b2838049cc bench : handle decode errors (#13548) b5391 Georgi Gerganov 2025-05-15 05:57:02 +03:00
  • aa48e373f2 server: inject date_string in llama 3.x template + fix date for firefunction v2 (#12802) b5390 Olivier Chafik 2025-05-15 02:39:51 +01:00
  • e3a9421b78 kv-cache : fix out-of-bounds view during reserve graph (#13547) Georgi Gerganov 2025-05-14 23:15:15 +03:00
  • 5ab5d5fb25 arm64: optimize q6_k_q8_k kernel with i8mm (#13519) b5388 Yibo Cai 2025-05-15 03:53:52 +08:00
  • 8282d74692 bench : handle decode errors gg/bench-handle-decode-errors Georgi Gerganov 2025-05-14 22:36:29 +03:00
  • 3198405e98 common: add partial regex support (#12808) b5387 Olivier Chafik 2025-05-14 19:50:57 +01:00
  • f5170c1d7a editorconfig : fix trailing whitespace from #13542 (#13546) Sigbjørn Skjæret 2025-05-14 20:22:49 +02:00
  • 017f10b5fa fix: crash when calling llama_state_get_size on a context without a KV cache (#13542) b5385 Gilad S. 2025-05-14 19:18:18 +03:00
  • 4696d56749 CUDA: fix crash on large batch size for quant. MoE (#13537) b5384 Johannes Gäßler 2025-05-14 16:41:02 +02:00
  • b7d2672082 llama : fix quantize with dl backends (#13539) Diego Devesa 2025-05-14 07:12:36 -07:00
  • 6da34fa276 CUDA: faster Deepseek FA, add Turing support (#13435) b5382 Johannes Gäßler 2025-05-14 16:08:20 +02:00
  • 5e7d95e22e fix: Move build_inp_pos to the top of the graph section for build_granite (#13538) b5381 Gabe Goodhart 2025-05-14 06:53:59 -06:00
  • 053174436f server : passthrough the /models endpoint during loading (#13535) b5380 Georgi Gerganov 2025-05-14 15:42:10 +03:00
  • 237acc7cd5 server : update readme + return json for "meta" field gg/server-models-loading Georgi Gerganov 2025-05-14 15:30:12 +03:00
  • 360a9c98e1 server : fix cache_tokens bug with no cache_prompt (#13533) b5379 Xuan-Son Nguyen 2025-05-14 13:35:07 +02:00
  • 6190e1c1c9 server : passthrough the /models endpoint during loading Georgi Gerganov 2025-05-14 14:15:42 +03:00
  • 09d13d94fb cmake: simplify vulkan shader test logic (#13263) b5378 bandoti 2025-05-14 07:53:57 -03:00
  • 24e86cae72 vulkan: KHR_coopmat flash attention (#13506) b5377 Jeff Bolz 2025-05-14 18:55:26 +09:00
  • bb1681fbd5 webui : use fflate for more deterministic gzip compress (#13525) Xuan-Son Nguyen 2025-05-14 10:26:12 +02:00
  • d486dd3e8e webui: Allow pasting file from clipboard (#13526) Luca Stefani 2025-05-14 10:07:31 +02:00
  • 21ca987fba docs: Update link to ggml-org in multimodal.md (#13513) ddpasa 2025-05-14 09:59:12 +02:00
  • be1d4a13db scripts : fix compare-llama-bench.py show parameter (#13514) Sigbjørn Skjæret 2025-05-14 08:41:01 +02:00
  • ab3971f2a0 vulkan: workaround FA compile failures on macos (#13517) b5372 Jeff Bolz 2025-05-14 13:15:50 +09:00
  • e5c834f718 quantize : improve tensor-type pattern matching (#13033) b5371 Ed Addario 2025-05-13 18:12:31 +01:00
  • 71bdbdb587 clip : clip.h become private API (⚠️ breaking change) (#13510) b5370 Xuan-Son Nguyen 2025-05-13 17:07:21 +02:00
  • f0995d28ce metal : use FA-vec kernel up to batch size 20 (#13496) b5369 Georgi Gerganov 2025-05-13 18:04:39 +03:00
  • c252e0c409 metal : optimize multi-sequence FA vec kernel (#13493) b5368 Georgi Gerganov 2025-05-13 18:04:00 +03:00
  • 4f711afed5 ggml-cpu: Update KleidiAI to v1.6 and fix include directives (#13509) b5367 Dan Johansson 2025-05-13 17:02:28 +02:00
  • b89d605a91 batched-bench : fix pp batch contents (#13492) b5366 Georgi Gerganov 2025-05-13 18:01:53 +03:00
  • b4726345ac mtmd : remove libllava, remove clip-quantize-cli (⚠️ breaking change) (#13460) b5365 Xuan-Son Nguyen 2025-05-13 15:33:58 +02:00
  • bf79371120 scripts : support arbitrary input file formats in compare-llama-bench.py (#13455) Sigbjørn Skjæret 2025-05-13 15:31:12 +02:00
  • d590cd4c24 model : Granite MoE shared (#13269) b5363 Gabe Goodhart 2025-05-13 07:12:01 -06:00
  • 1e2809bc4b sync : ggml Georgi Gerganov 2025-05-13 14:01:45 +03:00
  • 78d70223c3 metal : use FA-vec kernel up to batch size 20 gg/metal-fa-vec-bs20 Georgi Gerganov 2025-05-13 08:52:34 +03:00
  • fdfc7de7fc metal : optimize multi-sequence FA vec kernel Georgi Gerganov 2025-05-13 08:03:27 +03:00
  • f078c79865 batched-bench : fix pp batch contents Georgi Gerganov 2025-05-13 07:55:30 +03:00
  • cf0a43bb64 llama-bench : add defrag-thold, check for invalid ranges (#13487) b5361 Diego Devesa 2025-05-12 15:31:37 -07:00
  • f0d46ef157 opencl: remove unnecessary assert for add (#13257) b5360 lhez 2025-05-12 13:13:49 -07:00
  • de4c07f937 clip : cap max image size 1024 for qwen vl model (#13478) b5359 Xuan-Son Nguyen 2025-05-12 15:06:51 +02:00
  • 10d2af0eaa llama/ggml: add LLM training support (#10544) b5358 Johannes Gäßler 2025-05-12 14:44:49 +02:00
  • 064cc596ac context : fix state io for memory-less contexts (#13470) b5357 Georgi Gerganov 2025-05-12 15:12:27 +03:00
  • 91159ee9df server : allow content to be null in oaicompat_completion_params_parse (#13477) b5356 Anudit Nagar 2025-05-12 18:56:42 +07:00
  • 22cdab343b llama-bench : accept ranges for integer parameters (#13410) b5355 Diego Devesa 2025-05-12 13:08:22 +02:00
  • a71a4075cd ggml-cpu: Integrate fp32=bf16xbf16 SME KleidiAI kernel (#13053) b5354 Dan Johansson 2025-05-12 13:06:19 +02:00