upstream-archive/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-07-28 13:41:01 -05:00
Code Issues Packages Projects Releases Wiki Activity

Commit Graph

  • 141cab137d gguf-py : add MXFP4 de/quantization support Francis Couture-Harpin 2025-08-05 23:07:21 -04:00
  • 9515c6131a ggml: WebGPU disable SET_ROWS for now (#15078) b6097 Reese Levine 2025-08-05 16:26:38 -07:00
  • fd1234cb46 llama : add gpt-oss (#15091) b6096 Georgi Gerganov 2025-08-05 22:10:36 +03:00
  • f324a3b715 chat : only remove double bos/eos if added (#15086) b6095 Sigbjørn Skjæret 2025-08-05 20:43:36 +02:00
  • ea5e55d03e Merge branch 'master' into compilade/imatrix-neutral-prior Francis Couture-Harpin 2025-08-05 13:34:40 -04:00
  • 46a8601140 quantize : assume the neutral prior is equal imatrix weights Francis Couture-Harpin 2025-08-05 13:34:01 -04:00
  • be42642581 readme : update hot topics (#15097) Georgi Gerganov 2025-08-05 20:19:33 +03:00
  • 3306ceabf0 sycl: fix mul_mat selection (#15092) b6093 Romain Biessy 2025-08-05 18:39:55 +02:00
  • c81de6e107 Fix glm4moe bug (#15088) b6092 Juk Armstrong 2025-08-05 13:56:44 +01:00
  • 22f060c9c4 webui: fix markdown table (#15081) Alex Wu 2025-08-05 19:56:44 +08:00
  • ee3a9fcf88 context : fix index overflow on huge outputs (#15080) b6090 compilade 2025-08-05 05:27:45 -04:00
  • 2ec70c964b tests: Fix OPT_STEP_SGD test-backend-ops 0cc4m/vulkan-op-opt-step-sgd 0cc4m 2025-07-20 07:22:28 +00:00
  • 9d0312425e Vulkan: Implement GGML_OP_OPT_STEP_SGD 0cc4m 2025-07-20 07:18:30 +00:00
  • 145401c9e3 context : fix logits size overflow for huge batches compilade/fix-output-overflow Francis Couture-Harpin 2025-08-04 22:26:34 -04:00
  • f16a843a38 context : fix overflow when re-ordering huge outputs Francis Couture-Harpin 2025-08-04 22:01:28 -04:00
  • ec428b02c3 llama : add --n-cpu-moe option (#15077) b6089 Diego Devesa 2025-08-04 16:05:36 -07:00
  • 50e83eaed8 Merge branch 'master' into finelayer Jonathan Graehl 2025-08-04 15:44:10 -07:00
  • 19f68fa5a4 imatrix : warn when GGUF imatrix is saved without .gguf suffix (#15076) b6088 compilade 2025-08-04 17:26:52 -04:00
  • 41613437ff cmake: Add GGML_BACKEND_DIR option (#15074) b6087 Christian Kastner 2025-08-04 21:29:14 +02:00
  • 342e7014db imatrix : only warn about suffix when output format is unspecified compilade/imatrix-gguf-warning Francis Couture-Harpin 2025-08-03 18:12:06 -04:00
  • afa43e13c8 imatrix : add warning when suffix is not .gguf for GGUF imatrix Francis Couture-Harpin 2025-08-03 18:03:53 -04:00
  • e5bebe5251 gguf-py : add --chat-template-file to gguf_new_metadata (#15075) Sigbjørn Skjæret 2025-08-04 21:01:48 +02:00
  • ef0144c087 model: support GLM 4.5 family of models (#14939) b6085 Sam 2025-08-05 04:29:25 +10:00
  • 2721257e3e quantize : fix confusing error message if ftype is invalid (#15071) b6084 Sigbjørn Skjæret 2025-08-04 18:11:02 +02:00
  • 587d0118f5 ggml: WebGPU backend host improvements and style fixing (#14978) b6083 Reese Levine 2025-08-04 08:52:43 -07:00
  • 92383bfab3 quantize : store metadata for prior weight used for imatrix Francis Couture-Harpin 2025-08-04 01:47:00 -04:00
  • 5aa1105da2 vulkan: fix build when using glslang that does not support coopmat2 (#15062) b6082 Jeff Bolz 2025-08-04 00:09:19 -05:00
  • 0416ed2bb8 quantize : configurable neutral imatrix prior Francis Couture-Harpin 2025-08-03 16:28:37 -04:00
  • d31192b4ee imatrix : use GGUF by default (#14842) b6081 compilade 2025-08-03 16:00:05 -04:00
  • 0a2f5496be imatrix : fix 3d activation handling for hybrid and recurrent models (#14994) b6080 compilade 2025-08-03 15:49:13 -04:00
  • 11a3811164 memory : handle kv_unified for hybrid models (#15050) b6079 compilade 2025-08-03 15:43:07 -04:00
  • 97366dc6ab vocab : JetBrains Mellum pre-tokenizer (#15045) b6078 Csaba Kecskemeti 2025-08-03 12:38:18 -07:00
  • 83bc2f288c model : add text-only support for Kimi-VL (and find special tokens in text_config) (#15051) Gabriel Larson 2025-08-03 09:56:25 -05:00
  • 6c7a441161 vulkan: Use coopmat2 for conv2d (#14982) b6076 Jeff Bolz 2025-08-03 07:23:57 -05:00
  • e549515cb3 memory : handle kv_unified for hybrid models compilade/hybrid-kv_unified Francis Couture-Harpin 2025-08-03 00:45:47 -04:00
  • 5c0eb5ef54 opencl: fix adreno compiler detection logic (#15029) b6075 lhez 2025-08-02 10:51:18 -07:00
  • 03d4698218 CUDA: use mma FA kernel for gqa > 4 on RTX 4000 (#15035) b6074 Johannes Gäßler 2025-08-02 16:37:08 +02:00
  • 3303c19b16 cuda: make im2col a little faster (#15025) b6073 leejet 2025-08-02 22:15:36 +08:00
  • 4fdea540bd kv-cache : skip alignment of n_stream in kv-cache log msg [no ci] (#15040) Daniel Bevenius 2025-08-02 16:14:57 +02:00
  • a4569c41fd llama : enable LLAMA_SET_ROWS=1 by default (#14959) b6071 Georgi Gerganov 2025-08-02 17:14:21 +03:00
  • 15e92fd337 cuda, sycl : fix batched gemm when ne02 == 1 && ne03 > 1 (#15038) b6070 Georgi Gerganov 2025-08-02 17:13:05 +03:00
  • 2bf3fbf0b5 ci : check that pre-tokenizer hashes are up-to-date (#15032) Sigbjørn Skjæret 2025-08-02 14:39:01 +02:00
  • 711d5e6fe6 convert : fix Qwen3-Embedding pre-tokenizer hash (#15030) Douglas Hanley 2025-08-02 05:51:02 -05:00
  • f738989dcb chat : fix multiple tool_calls on hermes-2-pro (#14962) b6067 Jhen-Jie Hong 2025-08-02 18:04:48 +08:00
  • 4cb208c93c vulkan: coopmat2 mul_mat optimizations (#14934) b6066 Jeff Bolz 2025-08-02 04:21:37 -05:00
  • 3025b621d1 llama-bench: rename DB table name from test to llama_bench (#15003) b6065 R0CKSTAR 2025-08-02 17:20:40 +08:00
  • ec0b18802c vulkan: Support ne[3]>1 in noncontig matrix-vector multiply (#15015) b6064 Jeff Bolz 2025-08-02 03:48:30 -05:00
  • 339bd0268c model : support Qwen3-Embedding (#15023) b6063 Douglas Hanley 2025-08-02 03:44:50 -05:00
  • f906275537 server: enable token array inputs for OAI API (#15001) b6062 Johannes Gäßler 2025-08-02 10:12:41 +02:00
  • a9f7541ec2 vulkan: optimizations for direct convolution (#14933) b6061 Jeff Bolz 2025-08-02 02:57:04 -05:00
  • 9c35706b98 CUDA: fix MMQ nwarps for AMD with warp_size==32 (#15014) b6060 Johannes Gäßler 2025-08-01 20:47:32 +02:00
  • c76b420e4c vendor : update vendored copy of google/minja (#15011) b6059 l-austenfeld 2025-08-01 16:59:06 +02:00
  • 0f5ccd6fd1 model : add hunyuan dense (#14878) b6058 stevenkuang 2025-08-01 21:31:12 +08:00
  • 1c872f71fb opencl: add f16 for add, sub, mul, div (#14984) b6057 lhez 2025-08-01 04:15:44 -07:00
  • baad94885d ggml : Q2k interleaving implementation - x86/x64 SIMD (#14373) b6056 Srihari-mcw 2025-08-01 11:50:33 +05:30
  • ba42794c9e graph : fix equal_seq() check (#14986) b6055 Georgi Gerganov 2025-08-01 06:38:12 +03:00
  • 2860d479b4 docker : add cann build pipline (#14591) b6054 diannao 2025-08-01 10:02:34 +08:00
  • 484b2091ce compare-commits.sh: support both llama-bench and test-backend-ops (#14392) R0CKSTAR 2025-08-01 08:47:27 +08:00
  • daf2dd7880 quantize : skip tensor override when in fallback mode (#14995) b6052 Ed Addario 2025-07-31 20:32:18 +01:00
  • a06ed5feae llama : add simple option to enable CPU for MoE weights (--cpu-moe) (#14992) b6051 Diego Devesa 2025-07-31 11:15:41 -07:00
  • 784524053d Fix params bug in diffusion example (#14993) b6050 Aman Gupta 2025-08-01 01:22:58 +08:00
  • d6818d06a6 llama : allow other bufts when overriding to CPU, add --no-repack option (#14990) b6049 Diego Devesa 2025-07-31 09:11:34 -07:00
  • 91e67b8583 imatrix : fix 3d tensor counts compilade/imatrix-saner-3d Francis Couture-Harpin 2025-07-31 11:56:13 -04:00
  • e08a98826b Vulkan: Fix minor debug mode issues (#14899) b6048 Ruben Ortlam 2025-07-31 17:46:54 +02:00
  • 05beb070fc Merge branch 'master' into compilade/imatrix-saner-3d Francis Couture-Harpin 2025-07-31 11:25:26 -04:00
  • 952a47f455 mtmd : support MiniCPM-V 4.0 (#14983) b6047 tc-mb 2025-07-31 23:22:17 +08:00
  • d4f36e5e2b imatrix : fix 3d activations when model tensor is 2d Francis Couture-Harpin 2025-07-31 11:20:58 -04:00
  • 36e5fe7bcd MODEL_TENSOR.SSM_DT_NORM has defined twice (#14991) Csaba Kecskemeti 2025-07-31 07:59:49 -07:00
  • 94933c8c2e server : implement universal assisted decoding (#12635) b6045 g2mt 2025-07-31 05:25:23 -07:00
  • c1dacaa99b llama : merge build_moe_ffn_from_probs function into build_moe_ffn (#14968) b6044 Dongliang Wei 2025-07-31 20:12:20 +08:00
  • a9f77a8be3 server : add openai-style logit_bias support (#14946) b6043 Lukas Straub 2025-07-31 14:08:23 +02:00
  • 8a4a856277 Add LLaDA 8b Diffusion model (#14771) b6042 Aman Gupta 2025-07-31 19:49:09 +08:00
  • 11490b3672 CANN: Improve loading efficiency after converting weights to NZ format. (#14985) b6041 hipudding 2025-07-31 19:47:20 +08:00
  • 66625a59a5 graph : reduce splits for recurrent and hybrid models (#14825) b6040 compilade 2025-07-31 01:02:46 -04:00
  • 6e6725459a opencl: add mul_mat_f32_f32_l4_lm and mul_mat_f16_f32_l4_lm (#14809) b6039 lhez 2025-07-30 14:56:55 -07:00
  • e9192bec56 quantize : fix using combined imatrix GGUFs (multiple datasets) (#14973) b6038 Ed Addario 2025-07-30 20:11:56 +01:00
  • 41e78c567e server : add support for embd_normalize parameter (#14964) b6037 Daniel Bevenius 2025-07-30 18:07:11 +02:00
  • ad4a700117 HIP: enable mfma mmq on gfx908 and gfx90a for select datatypes and shapes (#14949) b6036 uvos 2025-07-30 17:38:06 +02:00
  • e32a4ec60e sync : ggml b6035 Georgi Gerganov 2025-07-30 16:03:13 +03:00
  • e228de9449 cmake : Fix BLAS link interface (ggml/1316) Kai Pastor 2025-07-30 14:53:16 +02:00
  • 73a8e5ca03 vulkan : fix 32-bit builds (ggml/1313) Kai Pastor 2025-07-30 14:52:26 +02:00
  • 92b8810ec7 CUDA: skip masked KV slices for all FA kernels (#14924) b6032 Johannes Gäßler 2025-07-30 15:46:13 +02:00
  • 00131d6eaf tests : update for LLAMA_SET_ROWS=1 (#14961) b6031 Georgi Gerganov 2025-07-30 15:12:02 +03:00
  • 1e15bfd42c graph : fix stack-use-after-return (#14960) b6030 Georgi Gerganov 2025-07-30 13:52:11 +03:00
  • a118d80233 embeddings: fix extraction of CLS pooling results (#14927) b6029 Douglas Hanley 2025-07-30 00:25:05 -05:00
  • 61550f8231 CANN: update ops docs (#14935) Xinpeng Dou 2025-07-30 08:39:24 +08:00
  • aa79524c51 HIP: remove the use of __HIP_PLATFORM_AMD__, explicitly support only AMD targets (#14945) b6027 uvos 2025-07-29 20:23:04 +02:00
  • b98f80a6b4 server : test alternative LRU logic gg/server-test-lru Georgi Gerganov 2025-07-29 21:19:21 +03:00
  • b77d11179d HIP: add GGML_HIP_MMQ_MFMA option to allow disableing the MFMA path. (#14930) b6026 uvos 2025-07-29 17:44:30 +02:00
  • c7aa1364fd HIP: Ignore unsupported unroll transformation in fattn-vec (#14931) b6025 uvos 2025-07-29 17:43:43 +02:00
  • 1a67fcc306 common : avoid logging partial messages (which can contain broken UTF-8 sequences) (#14937) b6024 kallewoof 2025-07-30 00:05:38 +09:00
  • 204f2cf168 CANN: Add ggml_set_rows (#14943) b6023 hipudding 2025-07-29 22:36:43 +08:00
  • 138b288b59 cuda : add softcap fusion (#14907) b6022 Sigbjørn Skjæret 2025-07-29 14:22:03 +02:00
  • 0591b39e48 ops: add MUSA xd/ops-musa Xiaodong Ye 2025-07-29 17:25:32 +08:00
  • bbd0f91779 server-bench: make seed choice configurable (#14929) Johannes Gäßler 2025-07-29 10:40:50 +02:00
  • 0a5036bee9 CUDA: add roll (#14919) b6020 Aman Gupta 2025-07-29 14:45:18 +08:00
  • 381879e0ac cont : tmp gg/repack-opt-mm-id Georgi Gerganov 2025-07-29 07:42:55 +03:00
  • fb371c18ec bench,common : add CPU extra buffer types gg/ot-cpu-repack Georgi Gerganov 2025-07-28 21:53:18 +03:00
  • 8ad7b3e65b opencl : add ops docs (#14910) lhez 2025-07-28 09:50:17 -07:00
  • bda62193b2 test-backend-ops : extend test case filtering (#14865) b6018 Leonard Mosescu 2025-07-28 09:04:27 -07:00