upstream-archive/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-07-29 06:00:58 -05:00
Code Issues Packages Projects Releases Wiki Activity

Commit Graph

  • 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
  • 95e18884fc CUDA: fix misaligned synchronization in FA (#13469) b5353 Johannes Gäßler 2025-05-12 10:51:21 +02:00
  • df8491922f ggml : add mrope kernel for metal (#13457) b5352 Xuan-Son Nguyen 2025-05-12 10:29:13 +02:00
  • 14492144c2 enable dpcpp nightly builds with libraries (#13406) b5351 Atharva Dubey 2025-05-12 06:15:32 +01:00
  • c104023994 mtmd : Use RMS norm for InternVL 3 38B and 78B mmproj (#13459) b5350 City 2025-05-12 00:39:06 +02:00
  • 9a390c4829 tools : fix uninitialized llama_batch in server (#13436) b5349 Anthony Umfer 2025-05-11 11:08:26 -04:00
  • 09232370fc scripts : exit compare-llama-bench.py gracefully when there's nothing to compare (#13451) Sigbjørn Skjæret 2025-05-11 16:20:39 +02:00
  • 7474e00b34 CUDA: fix crash with partial offloading of MoE (#13439) b5347 Johannes Gäßler 2025-05-11 16:09:33 +02:00
  • 7f323a589f Add --no-op-offload to improve -ot pp perf in MoE models like llama4 400B (#13386) b5346 David Huang 2025-05-11 20:18:39 +08:00
  • 3eac209319 mtmd : support InternVL 3 38B and 78B mmproj (#13443) b5345 City 2025-05-11 11:35:52 +02:00
  • a634d75d1b mtmd : move helpers to dedicated file (#13442) b5344 Xuan-Son Nguyen 2025-05-11 11:34:23 +02:00
  • 62d4250e52 docs : Fix typo in InternVL3 model name (#13440) Thomas Germer 2025-05-10 22:26:46 +02:00
  • 0208355f42 CUDA: fix race conditions FlashAttention kernels (#13438) b5342 Johannes Gäßler 2025-05-10 22:22:48 +02:00
  • d2a4ef05c6 vocab : add ByteDance-Seed/Seed-Coder (#13423) b5341 Sigbjørn Skjæret 2025-05-10 22:08:07 +02:00
  • 15e6125a39 mtmd : add hard limit on image resolution for qwen2vl / qwen2.5vl (#13434) b5340 Xuan-Son Nguyen 2025-05-10 19:57:54 +02:00
  • 3b24d26c22 server : update docs (#13432) Xuan-Son Nguyen 2025-05-10 18:44:49 +02:00
  • 43dfd741a5 llguidance : set tokenizer slices to default (#13424) b5338 Sigbjørn Skjæret 2025-05-10 17:19:52 +02:00
  • b064a51a4e ci: free_disk_space flag enabled for intel variant (#13426) Thammachart Chinvarapon 2025-05-10 21:34:48 +07:00
  • 053367d149 mtmd : support InternVL 2.5 and 3 (#13422) b5336 Xuan-Son Nguyen 2025-05-10 16:26:42 +02:00
  • d8919424f1 CUDA: fix FlashAttention on Turing (#13415) b5335 Johannes Gäßler 2025-05-10 09:16:52 +02:00
  • 7fef11766c arg : add env var to control mmproj (#13416) b5334 Xuan-Son Nguyen 2025-05-10 08:16:29 +02:00
  • dc1d2adfc0 vulkan: scalar flash attention implementation (#13324) b5333 Jeff Bolz 2025-05-09 23:07:07 -07:00
  • 7c28a74e07 chore(llguidance): use tagged version that does not break the build (#13413) b5332 Helton Reis 2025-05-09 17:15:39 -03:00
  • 33eff40240 server : vision support via libmtmd (#12898) b5331 Xuan-Son Nguyen 2025-05-09 19:29:37 +02:00
  • 17512a94d6 sycl : implementation of reordered Q4_0 MMVQ for Intel GPUs (#12858) b5330 Alberto Cabrera Pérez 2025-05-09 16:34:08 +01:00
  • 611aa914ef metal : optimize MoE for large batches (#13388) b5329 Georgi Gerganov 2025-05-09 15:14:56 +03:00
  • 0cf6725e9f CUDA: FA support for Deepseek (Ampere or newer) (#13306) b5328 Johannes Gäßler 2025-05-09 13:34:58 +02:00
  • 27ebfcacba llama : do not crash if there is no CPU backend (#13395) b5327 Diego Devesa 2025-05-09 13:02:07 +02:00
  • 5c86c9ed3e CUDA: fix crash on large batch size for MoE models (#13384) b5326 Johannes Gäßler 2025-05-09 12:14:04 +02:00
  • efb8b47eda imatrix : Add --parse-special for enabling parsing of special tokens in imatrix calculation (#13389) b5325 Bartowski 2025-05-09 05:53:58 -04:00
  • 0527771dd8 llama-run: add support for downloading models from ModelScope (#13370) b5324 R0CKSTAR 2025-05-09 17:25:50 +08:00
  • 2189fd3b63 mtmd : fix batch_view for m-rope (#13397) b5323 Xuan-Son Nguyen 2025-05-09 11:18:02 +02:00
  • 3f96aeff39 llama : one-off chat template fix for Mistral-Small-2503 (#13398) b5322 Xuan-Son Nguyen 2025-05-09 11:17:51 +02:00
  • b486ba05bf rpc : add rpc_msg_set_tensor_hash_req (#13353) b5321 Radoslav Gerganov 2025-05-09 10:31:07 +03:00
  • 02115dcd9a vulkan: Allow up to 4096 elements for mul_mat_id row_ids (#13326) b5320 Jeff Bolz 2025-05-09 02:23:41 -05:00
  • d9c4accaff server : (webui) rename has_multimodal --> modalities (#13393) Xuan-Son Nguyen 2025-05-09 09:06:37 +02:00
  • 15e03282bb ci : limit write permission to only the release step + fixes (#13392) b5318 Diego Devesa 2025-05-08 23:45:22 +02:00
  • f05a6d71a0 mtmd : Expose helper_decode_image_chunk (#13366) b5317 Matt Clayton 2025-05-08 14:25:39 -04:00
  • ee01d71e58 server : (webui) fix a very small misalignment (#13387) Xuan-Son Nguyen 2025-05-08 18:51:45 +02:00
  • 8c83449cb7 server : (webui) revamp the input area, plus many small UI improvements (#13365) b5315 Xuan-Son Nguyen 2025-05-08 15:37:29 +02:00
  • 1a844be132 convert : support rope_scaling type and rope_type (#13349) Sigbjørn Skjæret 2025-05-08 15:34:29 +02:00
  • 0ccc121354 mtmd : fix the calculation of n_tokens for smolvlm (#13381) b5313 welix 2025-05-08 22:03:53 +09:00
  • 6562e5a4d6 context : allow cache-less context for embeddings (#13108) Georgi Gerganov 2025-05-08 14:28:33 +03:00
  • 51fb96b1ff context : remove logits_all flag (#13284) b5311 Georgi Gerganov 2025-05-08 14:26:50 +03:00
  • 70a6991edf ci : move release workflow to a separate file (#13362) b5310 Diego Devesa 2025-05-08 13:15:28 +02:00
  • f061021206 llama : print size and type of overridden tensors (#13364) b5309 Diego Devesa 2025-05-08 13:15:15 +02:00
  • 6107303ab0 llama : remove logits_all flag + reorder llama_context_params gg/context-remove-logits-all Georgi Gerganov 2025-05-08 12:56:39 +03:00
  • 6c0501adf7 context : remove logits_all flag Georgi Gerganov 2025-05-03 19:21:10 +03:00
  • 8733e0cf6e sycl: addressing non-contiguous src1 mul_mats (nc and batched) (#13343) b5308 Alberto Cabrera Pérez 2025-05-08 10:08:01 +01:00
  • 814f795e06 docker : disable arm64 and intel images (#13356) Diego Devesa 2025-05-07 16:36:33 +02:00
  • d879433824 sync : ggml b5306 Georgi Gerganov 2025-05-07 16:39:36 +03:00
  • 13b0a04597 whisper: remove MSVC warnings pragmas (whisper/3090) Daniel Bevenius 2025-05-05 13:09:35 +02:00
  • bba9d945c1 cmake : removed stdc++fs (whisper/3097) Jared Tweed 2025-05-02 02:41:35 -07:00
  • bc4e1128f7 llama : deci : support ffn-free with attention (#13296) b5303 Sigbjørn Skjæret 2025-05-07 12:49:27 +02:00
  • 39e73ae0d6 common : Add a warning when we can't match samplers from a string or char. (#13330) b5302 Ycros 2025-05-07 18:23:28 +10:00
  • 1f73301b63 cuda : remove nrows_x in mul_mat_q_process_tile (#13325) b5301 R0CKSTAR 2025-05-07 15:48:23 +08:00
  • 4773d7a02f examples : remove infill (#13283) b5300 Georgi Gerganov 2025-05-07 10:28:02 +03:00
  • 6c7fd67b64 llama : support tie embedding for chatglm models (#13328) b5299 piDack 2025-05-07 15:23:11 +08:00
  • 141a908a59 CUDA: mix virt/real CUDA archs for GGML_NATIVE=OFF (#13135) b5298 Johannes Gäßler 2025-05-06 23:35:51 +02:00
  • 32916a4907 clip : refactor graph builder (#13321) b5297 Xuan-Son Nguyen 2025-05-06 22:40:24 +02:00
  • ffc727203a sampling : make top_n_sigma no-op at <=0 or a single candidate (#13345) b5296 DocShotgun 2025-05-06 13:36:24 -07:00
  • 91a86a6f35 sampling : don't consider -infinity values in top_n_sigma (#13344) b5295 oobabooga 2025-05-06 15:24:15 -03:00
  • f4ed10b69c cmake : remove arm64 msvc presets (#13342) Diego Devesa 2025-05-06 20:15:31 +02:00
  • 1e333d5bba SYCL: Disable reorder optimize by default and stop setting tensor extras when optimize is disabled (#13254) b5293 Akarshan Biswas 2025-05-06 20:27:06 +05:30
  • 2f54e348ad llama : fix build_ffn without gate (#13336) b5292 Xuan-Son Nguyen 2025-05-06 14:25:40 +02:00
  • 2356fb1d53 CUDA: fix bad asserts for partial offload (#13337) Johannes Gäßler 2025-05-06 13:58:51 +02:00
  • 764b85627b convert : qwen2/3moe : set yarn metadata if present (#13331) Sigbjørn Skjæret 2025-05-06 11:12:06 +02:00
  • 15a28ec8c7 CUDA: fix --split-mode row for MMQ (#13323) b5289 Johannes Gäßler 2025-05-06 08:36:46 +02:00
  • a7366faa5b gguf-py : avoid requiring pyside6 for other scripts (#13036) gguf-v0.16.3 compilade 2025-05-05 22:27:31 -04:00
  • 9070365020 CUDA: fix logic for clearing padding with -ngl 0 (#13320) b5287 Johannes Gäßler 2025-05-05 22:32:13 +02:00
  • 233461f812 sampling : Integrate Top-nσ into main sampling chain (and add it to the server) (#13264) b5286 oobabooga 2025-05-05 17:12:19 -03:00
  • b34c859146 server : Webui - change setText command from parent window to also send the message. (#13309) igardev 2025-05-05 17:03:31 +03:00
  • 9b61acf060 mtmd : rename llava directory to mtmd (#13311) b5284 Xuan-Son Nguyen 2025-05-05 16:02:55 +02:00
  • 5215b91e93 clip : fix confused naming ffn_up and ffn_down (#13290) b5283 Xuan-Son Nguyen 2025-05-05 12:54:44 +02:00
  • ae803bfc3d convert : bailingmoe : set yarn metadata if present (#13312) Sigbjørn Skjæret 2025-05-05 12:34:26 +02:00
  • 66645a5285 SYCL: Disable mul_mat kernels for noncontiguous tensor b (#13308) b5281 Akarshan Biswas 2025-05-05 13:39:10 +05:30
  • 27aa259532 mtmd : add C public API (#13184) b5280 Xuan-Son Nguyen 2025-05-04 23:43:42 +02:00
  • 9fdfcdaedd rpc : use backend registry, support dl backends (#13304) b5279 Diego Devesa 2025-05-04 21:25:43 +02:00
  • 6eb7d25c70 ggml : activate s390x simd for Q3_K (#13301) b5278 Aaron Teo 2025-05-05 01:49:12 +08:00
  • 86bd60d3fe llava/mtmd : fixes to fully support dl backends (#13303) b5277 Diego Devesa 2025-05-04 17:05:20 +02:00
  • 9f2da5871f llama : build windows releases with dl backends (#13220) b5276 Diego Devesa 2025-05-04 14:20:49 +02:00
  • 93c4e23905 CUDA: fix race condition in MMQ stream-k fixup (#13299) b5275 Johannes Gäßler 2025-05-04 14:16:39 +02:00
  • 8afbd96818 CUDA: fix race condition in MMQ ids_dst (#13294) b5274 Johannes Gäßler 2025-05-04 13:58:38 +02:00
  • 16843dba33 metal : pad mm results gg/metal-mm-pad Georgi Gerganov 2025-05-04 09:13:52 +03:00
  • 8ae5ebcf85 vulkan: Additional type support for unary, binary, and copy (#13266) b5273 Jeff Bolz 2025-05-04 00:17:16 -05:00
  • 3e959f0976 imatrix: fix oob writes if src1 is not contiguous (#13286) b5272 Johannes Gäßler 2025-05-04 00:50:37 +02:00
  • 36667c8edc clip : revert the change of BOI/EOI token for GLM-edge (⚠️ breaking change) (#13259) b5271 Xuan-Son Nguyen 2025-05-03 20:07:54 +02:00
  • 3bf785f3ef llama : Llama-3_1-Nemotron-Ultra-253B-v1 support (#12843) b5270 ymcki 2025-05-03 23:39:51 +08:00
  • e94f3932f2 kv-cache : allow context shift for recurrent models Francis Couture-Harpin 2025-05-02 19:29:23 -04:00
  • d55b0d0621 convert : avoid AutoConfig for Mamba and Mamba2 hparams Francis Couture-Harpin 2025-05-02 18:24:55 -04:00
  • 1d36b3670b llama : move end-user examples to tools directory (#13249) b5269 Diego Devesa 2025-05-02 20:27:13 +02:00
  • 15dea7bbdf opt : remove print [no ci] jg/llama-opt-3 Georgi Gerganov 2025-04-25 11:55:49 +03:00
  • cee751c450 opt : fix n_outputs Georgi Gerganov 2025-04-25 11:45:21 +03:00
  • 4e73b81a67 try CI fix Johannes Gäßler 2025-01-27 18:33:34 +01:00
  • 111c9c75d6 llama/ggml: add LLM training support Johannes Gäßler 2024-11-17 14:58:51 +01:00
  • b34443923c sync : ggml (#13268) Georgi Gerganov 2025-05-02 20:54:30 +03:00
  • a75cb30dc9 context : fix reorder logic (#13267) b5267 Georgi Gerganov 2025-05-02 20:54:13 +03:00