Commit Graph

  • d6cfdc669c ggml-webgpu: move from parameter buffer pool to single buffer with offsets (llama/21278) Reese Levine 2026-04-03 11:40:14 -07:00
  • 3f5117610b ggml-zendnn : add MUL_MAT_ID op support for MoE models (llama/21315) Vishal Singh 2026-04-03 14:49:08 +05:30
  • 321f628239 rpc : reuse compute graph buffers (llama/21299) Radoslav Gerganov 2026-04-03 10:28:09 +03:00
  • c5a5e6528e ggml-webgpu: add vectorized flash attention (llama/20709) Zheyuan Chen 2026-04-02 10:40:42 -07:00
  • 7f6c0ac20f sycl : fix llama_kv_cache hang when kv_cache is huge: 5GB (llama/21283) Neo Zhang 2026-04-02 15:08:32 +08:00
  • 514eabc1e5 ggml : bump version to 0.9.11 (ggml/1456) Georgi Gerganov 2026-04-02 10:37:26 +03:00
  • 444662bc83 hexagon : add cumsum op support (llama/21246) Todor Boinovski 2026-04-01 17:44:02 -07:00
  • 08108512c7 opencl: fix leak in Adreno q8_0 path (llama/21212) lhez 2026-04-01 12:54:58 -07:00
  • 82bb26fba1 CUDA: fix FA kernel selection logic (llama/21271) Johannes Gäßler 2026-04-01 21:28:19 +02:00
  • 9a40dd9365 hexagon: improve RMS_NORM and DIV accuracy (llama/21251) Aparna M P 2026-04-01 21:13:08 +05:30
  • fab70d287e sycl : support nvfp4 type in mul_mat (llama/21227) Neo Zhang 2026-04-01 18:54:15 +08:00
  • 981195be5a ggml-cuda: Add generic NVFP4 MMQ kernel (llama/21074) Michael Wand 2026-04-01 03:04:58 -07:00
  • ace95aac6b ggml : bump version to 0.9.10 (ggml/1454) Georgi Gerganov 2026-04-01 16:01:45 +03:00
  • 1971a362dc CUDA/HIP: Fix kernel slection for mmvq mmid kernel to align host selection with device launch bounds (llama/21238) uvos 2026-04-01 10:21:20 +02:00
  • 5c5b88eb77 ggml : fix RWKV ops thread assignment (llama/21226) Georgi Gerganov 2026-04-01 11:10:25 +03:00
  • 1b95f84550 ggml-cpu: fix fallback for RVV kernels without zvfh (llama/21157) Taimur Ahmad 2026-04-01 13:10:03 +05:00
  • 933bd1f79c CUDA: Add Flash Attention Support for Head Dimension 512 (llama/20998) Anav Prasad 2026-04-01 07:07:24 +00:00
  • 78f54d15d8 ggml webgpu: quantized buffers to u32 + wider browser/device support (llama/21046) Reese Levine 2026-03-31 22:38:24 -07:00
  • 21b9dd6789 ggml-webgpu: port all AOT operators to JIT (llama/20728) Abhijit Ramesh 2026-04-01 12:58:53 +03:00
  • 5ffe58838d CANN: fix multi-thread set_tensor race conditions (llama/20151) hipudding 2026-03-31 22:00:51 +08:00
  • 952c66237d sycl : enhance fattn perf (llama/21185) Neo Zhang 2026-03-31 18:31:50 +08:00
  • 6ac5a50005 opencl: add q4_K gemm and gemv kernels for Adreno (llama/20919) shaofeiqi 2026-03-30 12:19:16 -07:00
  • 75b9543856 CUDA : Fix CUB's argsort when nrows % block_size == 0 CCCL < 3.1 (llama/21181) Oliver Simons 2026-03-30 16:20:00 +02:00
  • 40ddc5a5b9 rpc : fix misleading error log (llama/21184) Radoslav Gerganov 2026-03-30 17:05:11 +03:00
  • 6b67c91879 Optimize MOE GEMV kernel for BS > 1. (llama/20905) Gaurav Garg 2026-03-29 22:05:18 +05:30
  • 9e96d390f7 hexagon: dma optimizations (mostly fixing regressions) (llama/21137) Max Krasnyansky 2026-03-29 06:40:13 -07:00
  • 763a454052 ggml : bump version to 0.9.9 (ggml/1449) Georgi Gerganov 2026-03-30 18:34:29 +03:00
  • fc674574ca bench : sync submit-results URL to ggml-org (#3769) jinweihan 2026-04-19 22:12:57 -07:00
  • 166c20b473 whisper : add stateless VAD detect + explicit state reset for streaming (#3677) Daniel Worthington-Bodart 2026-04-17 12:36:27 +01:00
  • 95ea8f9bfb sync : ggml Georgi Gerganov 2026-03-29 13:23:24 +03:00
  • 759f0084b4 vulkan: add noncontiguous GLU support (llama/21081) Ruben Ortlam 2026-03-28 08:44:56 +01:00
  • 52699f6d19 hexagon: support for IQ4_NL and MXFP4 (llama/21018) Yiwei Shao 2026-03-27 09:22:41 -07:00
  • 7f466e237b rpc : proper handling of data pointers to CPU buffers (llama/21030) Radoslav Gerganov 2026-03-27 10:59:35 +02:00
  • b564a99ed6 metal : Fix dimension constraint violation in matmul2d descriptor (llama/21048) ren 2026-03-27 00:05:21 -07:00
  • 45a7083431 hip: use fnuz fp8 for conversion on CDNA3 (llama/21040) uvos 2026-03-26 23:06:33 +01:00
  • 1848f994e3 opencl: allow large buffer for adreno (llama/20997) lhez 2026-03-26 08:52:21 -07:00
  • 07237ff99e fix(ggml): correct RISC-V ISA string canonical ordering for RVV in CMake (llama/20888) ihb2032 2026-03-26 19:08:41 +08:00
  • eb747f3def ggml-cuda: Add NVFP4 dp4a kernel (llama/20644) Michael Wand 2026-03-26 01:54:03 -07:00
  • a050c7d1bf CUDA & CPU: support F32 kernel type for CONV_TRANSPOSE_2D (llama/17094) Yihao Wang 2026-03-25 19:19:14 -07:00
  • 495b77aec2 mtmd: Add DeepSeekOCR Support (llama/17400) Saba Fallah 2026-03-25 19:57:40 +01:00
  • 3987857d2d llama: fix llama-model-saver (llama/20503) Johannes Gäßler 2026-03-25 11:53:16 +01:00
  • f2a8e65ea7 sycl : fix wrong variable check by assert (llama/20903) Neo Zhang 2026-03-25 17:48:37 +08:00
  • 9e4e4c2401 metal : add FLOOR, CEIL, ROUND, TRUNC unary ops (llama/20930) nuri 2026-03-24 17:13:07 +09:00
  • eef7422d4d metal : add FA instantiations for HSK=512, HSV=512 (llama/20902) Georgi Gerganov 2026-03-24 10:03:09 +02:00
  • 116a9f6ab7 hexagon: general DMA and Binary Op fixes for large strides (llama/20918) Max Krasnyansky 2026-03-23 15:33:49 -07:00
  • 624be93425 opencl: add q6_K gemm and gemv kernels for Adreno (llama/20089) lhez 2026-03-23 12:44:18 -07:00
  • 37c0a52c1b rpc : RCE patch (llama/20908) las7 2026-03-23 10:54:57 -07:00
  • c589dd77d4 metal: add CONV_3D (llama/19927) Rashid Ul Islam 2026-03-23 13:15:34 +05:30
  • 54f5c02f29 CANN: add RoPE cache preload before ACL graph capture (llama/20747) Chenguang Li 2026-03-23 15:24:06 +08:00
  • a0e41ec261 fix(openvino): explicit memset in buffer_context allocation (llama/20857) Dan Hoffman 2026-03-22 23:05:37 -07:00
  • c976b22d7b opencl: add flattened Q4_K mv and general Q4_K mm (llama/20773) shaofeiqi 2026-03-22 22:45:11 -07:00
  • 607c92430f CUDA: fix BF16 FA compilation (llama/20865) Johannes Gäßler 2026-03-22 17:53:33 +01:00
  • 1d0f0285de support bf16 and quantized type (llama/20803) Neo Zhang 2026-03-22 22:06:27 +08:00
  • 69f0d907ee ggml-cuda: native bf16 flash attention for vec kernel (llama/20525) Patrick Buckley 2026-03-22 03:05:51 -07:00
  • 77b635e9c4 Increase number of output elements per-thread block if the K-dimension is small (llama/20635) Gaurav Garg 2026-03-22 14:19:35 +05:30
  • 5f3428219a fix(rpc): prevent division by zero in deserialize_tensor (llama/20712) y198 2026-03-21 20:59:43 +07:00
  • 22710fdb82 Add shader count for Intel Arc Pro B60 (llama/20818) Matt Corallo 2026-03-21 04:22:51 +00:00
  • ca5d565dcd ggml-cpu: add always_inline to tinyBLAS_PPC accumulator saves (llama/20791) shalinib-ibm 2026-03-21 04:41:45 +05:30
  • 49b505bcc5 vulkan: change gated_delta_net to shard a column across a subgroup (llama/20662) Jeff Bolz 2026-03-20 06:17:15 -05:00
  • 46dcb35aa3 CANN: add BF16 support for core operators (llama/20152) hipudding 2026-03-20 17:08:39 +08:00
  • 65d820a44a ggml: guard KleidiAI DOWNLOAD_EXTRACT_TIMESTAMP for cmake < 3.24 (llama/20767) Sundaram krishnan 2026-03-20 01:06:23 +05:30
  • e1cdce46c5 hip: Avoid compiler bug in RDNA code generation during debug builds on Windows (llama/20655) Rail Chabdarov 2026-03-19 19:14:08 +01:00
  • 15f6b6ad76 hexagon: add Matrix Extensions (HMX) for Hexagon NPU backend (llama/20693) Yiwei Shao 2026-03-19 09:11:06 -07:00
  • 081dc773a5 ci : add hip quality check (llama/20430) uvos 2026-03-19 17:05:44 +01:00
  • 551bb82960 ggml webgpu: ops support for qwen3.5 (SET, TRI_SOLVE, SSM_CONV, GATED_DELTA_NET) + GET_ROWS optimization (llama/20687) Reese Levine 2026-03-19 08:45:28 -07:00
  • 43c7c0f86c vulkan: dequantize iq4_xs 4 at a time (llama/20657) Eve 2026-03-19 10:32:04 +00:00
  • fea629d00f cmake : fix build warning when kleidiai is enabled (llama/20457) Charles Xu 2026-03-19 09:14:48 +01:00
  • 2a6de29364 CANN: handle in-place ROPE on non-contiguous f32 tensors (llama/20274) Chenguang Li 2026-03-19 14:05:01 +08:00
  • 3d004fbf0a ggml-webgpu: Update the RMS_NORM preprocessor and add L2_NORM (llama/20665) Masashi Yoshimura 2026-03-28 11:47:59 +02:00
  • 12015a2174 ggml-webgpu: Add supports for DIAG and TRI (llama/20664) Masashi Yoshimura 2026-03-19 13:08:35 +09:00
  • dfba84cb47 CANN: support flash attention for head dim not multiple of 16, fix ALiBi slope offset (llama/20031) Chenguang Li 2026-03-19 11:02:42 +08:00
  • d6a0f0d075 Move to no timeout for WaitAny in graph submission to avoid deadlocks in some cases on llvm-pipe backends (llama/20618) Reese Levine 2026-03-18 10:23:47 -07:00
  • 14caedfa18 ggml-cpu/x86: fix unused changemask warning in repack (llama/20692) Shaw Nguyen 2026-03-18 23:45:06 +07:00
  • 61c7cd024d HIP : ignore return of hipMemAdvise [no ci] (llama/20696) uvos 2026-03-18 09:53:13 +01:00
  • e222814fc4 hexagon: add neg, exp, sigmoid, softplus ops, cont, repeat ops (llama/20701) Krishna Sridhar 2026-03-17 15:34:36 -07:00
  • 16ca5e6fb1 vulkan: disable mmvq on Intel Windows driver (llama/20672) Ruben Ortlam 2026-03-17 21:51:43 +01:00
  • 906aef3da8 ggml-blas: set mkl threads from thread context (llama/20602) Kevin Hannon 2026-03-17 13:16:49 -04:00
  • c890a9d9b4 ggml-cpu: fix RVV checks in quants and repacking (llama/20682) Taimur Ahmad 2026-03-17 19:03:40 +05:00
  • 0ad6ceef59 vulkan: async and event fixes (llama/20518) Ruben Ortlam 2026-03-17 14:27:23 +01:00
  • ab7d305b75 kleidiai : fix MUL_MAT support for batched (3D) inputs (llama/20620) Justin Bradford 2026-03-17 05:03:54 -07:00
  • 49adc8b470 vulkan: allow graphics queue only through env var (llama/20599) Ruben Ortlam 2026-03-17 10:09:59 +01:00
  • 6494251197 ehance UPSCALE to support all UT cases (llama/20637) Neo Zhang 2026-03-17 10:01:52 +08:00
  • 9232af59ba kleidiai: add data type check to get_tensor_traits (llama/20639) Martin Klacer 2026-03-16 19:25:54 +00:00
  • 724ea71cf9 vulkan: fix flash attention dot product precision (llama/20589) Ruben Ortlam 2026-03-16 10:45:49 +01:00
  • dae7781052 CUDA: GDN hide memory latency (llama/20537) Aman Gupta 2026-03-16 11:41:45 +08:00
  • 1335dfa785 sycl : fix for untransposed GDA recurrent state (llama/20583) Sigbjørn Skjæret 2026-03-15 19:10:15 +01:00
  • 76684141a5 ruby : fix dangling pointers, memory leak, and SEGV on parallel transcription (#3715) KITAITI Makoto 2026-03-22 02:03:00 +09:00
  • 9386f23940 release : v1.8.4 v1.8.4 Georgi Gerganov 2026-03-19 10:40:13 +02:00
  • ef3463bb29 ci : update workflows Georgi Gerganov 2026-03-18 22:43:38 +02:00
  • 4bbce1e5b2 benches : update gg/benches-update Georgi Gerganov 2026-03-18 22:34:51 +02:00
  • f5b477ab09 sync : ggml Georgi Gerganov 2026-03-18 14:45:25 +02:00
  • b2be16208d ggml : bump version to 0.9.8 (ggml/1442) Georgi Gerganov 2026-03-16 20:15:14 +02:00
  • 945d3151d9 ggml : restore ggml_type_sizef() to aboid major version bump (ggml/1441) Georgi Gerganov 2026-03-16 20:09:25 +02:00
  • dc96116622 fix: VAD time mapping timestamp drift caused by overlap samples (#3711) lohopupa 2026-03-17 12:19:08 +06:00
  • 79218f51d0 go : handle EOF correctly in model download (#3671) Alan 2026-03-16 12:44:18 +01:00
  • 975b979834 py : replace deprecated openvino-dev with openvino>=2023.3.0 (#3678) Aiudadadadf 2026-03-16 12:41:54 +01:00
  • 21665eab4c examples : Allow max_len to be used for any output format (#3679) Gaël James 2026-03-16 12:33:56 +01:00
  • 136dc2eb12 server: return proper HTTP status codes for error responses (#3707) Igor Loskutov 2026-03-16 07:33:06 -04:00
  • 27fa20774a ggml : try fix arm build (#0) Georgi Gerganov 2026-03-16 09:11:13 +02:00
  • 2bc630f197 talk-llama : sync llama.cpp Georgi Gerganov 2026-03-16 07:16:46 +02:00