| 2026-07-10 |
Ruben Ortlam | vulkan: for small AMD GPUs, reduce submission threshold... |
commit | commitdiff | tree |
| 2026-07-10 |
Max Krasnyansky | hexagon: new vtcm layouts and improved pipelines for... |
commit | commitdiff | tree |
| 2026-07-10 |
fairydreaming | cuda : add support for f16->f16 GGML_OP_SET_ROWS (llama... |
commit | commitdiff | tree |
| 2026-07-10 |
fairydreaming | metal : add set_rows with src0 f16 (llama/25434) |
commit | commitdiff | tree |
| 2026-07-10 |
tyronecai | ggml : fix A indexing in simd_gemm scalar tail-column... |
commit | commitdiff | tree |
| 2026-07-10 |
fairydreaming | ggml : add support for CPU f16->f16 GGML_OP_SET_ROWS... |
commit | commitdiff | tree |
| 2026-07-10 |
lhez | opencl: fix potential crash in aos reconstruct (llama... |
commit | commitdiff | tree |
| 2026-07-10 |
Pasha Khosravi | Add Q2_0 quantization: type definition and CPU backend... |
commit | commitdiff | tree |
| 2026-07-10 |
Oliver Simons | CUDA: Fuse MMVQ post-scale for NVFP4 (llama/24481) |
commit | commitdiff | tree |
| 2026-07-10 |
asf0 | ggml-hip : add -fno-finite-math-only alongside -ffast... |
commit | commitdiff | tree |
| 2026-07-10 |
Neo Zhang | fix unsupported UT cases of CONT & CPY (llama/25231) |
commit | commitdiff | tree |
| 2026-07-10 |
Neo Zhang | support op col2im_1d (llama/25264) |
commit | commitdiff | tree |
| 2026-07-10 |
Neo Zhang | support OP cross_entropy_loss, cross_entropy_loss_back... |
commit | commitdiff | tree |
| 2026-07-10 |
Todd Malsbary | sycl : set K_QUANTS_PER_ITERATION to 1 on DMMV path... |
commit | commitdiff | tree |
| 2026-07-10 |
Neo Zhang | fix unsupport ACC UT cases for noncontiguous (llama... |
commit | commitdiff | tree |
| 2026-07-10 |
Neo Zhang | sycl : enhance argsort to support all UT cases (llama... |
commit | commitdiff | tree |
| 2026-07-10 |
Neo Zhang | sycl : use sycl func to fix AOT double type issue ... |
commit | commitdiff | tree |
| 2026-07-10 |
Neo Zhang | sycl : rename the env vars from "disable" to "enable... |
commit | commitdiff | tree |
| 2026-07-10 |
An Long | ggml : make ggml_time_init idempotent (llama/24422) |
commit | commitdiff | tree |
| 2026-07-10 |
fairydreaming | vulkan : check src0 type in GGML_OP_SET_ROWS to avoid... |
commit | commitdiff | tree |
| 2026-07-10 |
Hongqiang Wang | opencl: general flash attention decode performance... |
commit | commitdiff | tree |
| 2026-07-10 |
Pascal | metal: add col2im_1d op (f32/f16/bf16) (llama/25176) |
commit | commitdiff | tree |
| 2026-07-10 |
Johannes Gäßler | CUDA: remove -sm row, refactor cuBLAS (llama/24216) |
commit | commitdiff | tree |
| 2026-07-10 |
Alexey Kopytko | CUDA: extend K-type validation to V-types for flash... |
commit | commitdiff | tree |
| 2026-07-10 |
ragz4125 | ggml-cpu: use UE4M3 LUT in ARM NVFP4 dot product (llama... |
commit | commitdiff | tree |
| 2026-07-10 |
shalinib-ibm | ggml-cpu: Enable tiled matmul on AIX (llama/25199) |
commit | commitdiff | tree |
| 2026-07-10 |
hokanosekai | vulkan: fix 32-bit integer overflow in CEIL_DIV (llama... |
commit | commitdiff | tree |
| 2026-07-10 |
a-huk | ggml-hip: enable -ffast-math for HIP builds (llama... |
commit | commitdiff | tree |
| 2026-07-10 |
adavyas | ggml-cuda: optimize conv_transpose_1d indexing (llama... |
commit | commitdiff | tree |
| 2026-07-10 |
Eve | abort if we see a multi buffer (llama/25276) |
commit | commitdiff | tree |
| 2026-07-10 |
liminfei-amd | ggml : fix tensor-parallel + -ncmoe crash on MoE models... |
commit | commitdiff | tree |
| 2026-07-10 |
Vexxie | ggml: Update VMM Pool allocation ggml-cuda.cu - Turing... |
commit | commitdiff | tree |
| 2026-07-10 |
fairydreaming | cuda : concat implementation for quantized types (llama... |
commit | commitdiff | tree |
| 2026-07-10 |
fairydreaming | ggml : fix broken CPU concat implementation for quantiz... |
commit | commitdiff | tree |
| 2026-07-10 |
Piotr Wilkin... | cuda: enable topk-moe fusion for 288 experts (llama... |
commit | commitdiff | tree |
| 2026-07-10 |
Gaurav Garg | Remove redundant CUDA copies after gated_delta_net... |
commit | commitdiff | tree |
| 2026-07-10 |
lhez | opencl: allow loading precompiled binary kernels from... |
commit | commitdiff | tree |
| 2026-07-10 |
Max Krasnyansky | hexagon: flash attention rework (optimizations, accurac... |
commit | commitdiff | tree |
| 2026-07-10 |
Johannes Gäßler | CUDA: consistent use of __restrict__ + PDL for FA ... |
commit | commitdiff | tree |
| 2026-07-10 |
ragz4125 | ggml-cpu: add AVX2 optimization for nvfp4 dot product... |
commit | commitdiff | tree |
| 2026-07-10 |
lhez | opencl: initial q1_0 support (llama/25160) |
commit | commitdiff | tree |
| 2026-07-10 |
fairydreaming | cuda : prevent integer truncation and overflow errors... |
commit | commitdiff | tree |
| 2026-07-10 |
Matt Jallo | CUDA: fix get_rows_back for tables with more than 65535... |
commit | commitdiff | tree |
| 2026-07-10 |
Johannes Gäßler | CUDA: fix Gemma E4B MTP FlashAttention (llama/25148) |
commit | commitdiff | tree |
| 2026-07-10 |
Kevin Liu | vulkan: roll bk loop in matmul for asahi linux (llama... |
commit | commitdiff | tree |
| 2026-07-10 |
zduford | HIP: use hipBLAS for dense prefill on gfx900, keep... |
commit | commitdiff | tree |
| 2026-07-10 |
Masashi Yoshimura | ggml-webgpu: add support for NVFP4 (llama/25143) |
commit | commitdiff | tree |
| 2026-07-10 |
Oliver Simons | Revert "sched : reintroduce less synchronizations durin... |
commit | commitdiff | tree |
| 2026-07-10 |
Ruben Ortlam | vulkan: use flops instead of weight tensor size for... |
commit | commitdiff | tree |
| 2026-07-10 |
Hongqiang Wang | opencl: flash attention improvement (llama/25069) |
commit | commitdiff | tree |
| 2026-07-10 |
Gaurav Garg | Added a cudaMemcpy2DAsync fast path to ggml_cuda_cpy... |
commit | commitdiff | tree |
| 2026-07-10 |
Neo Zhang | sycl : fix failed ut cases of norm (llama/25044) |
commit | commitdiff | tree |
| 2026-07-10 |
Ruben Ortlam | vulkan: fix step operator for 0 input (llama/25036) |
commit | commitdiff | tree |
| 2026-07-10 |
Andreas Kieslinger | sched : reintroduce less synchronizations during split... |
commit | commitdiff | tree |
| 2026-07-10 |
Ravi Panchumarthy | openvino: Update to OV 2026.2.1, self-contained release... |
commit | commitdiff | tree |
| 2026-07-10 |
nullname | vulkan: opt mul_mat_vecq for mi50 (llama/22933) |
commit | commitdiff | tree |
| 2026-07-10 |
Jiang, Fish | vulkan: add INTEL_XE1 arch enum and enable coopmat1... |
commit | commitdiff | tree |
| 2026-07-01 |
Lin Xiaodong | whisper : expose internal VAD speech segments (#3916) |
commit | commitdiff | tree |
| 2026-07-01 |
Daniel Bevenius | ci, examples : add package-lock.json to examples addon... |
commit | commitdiff | tree |
| 2026-07-01 |
Lin Xiaodong | whisper : map token timestamps to original time when... |
commit | commitdiff | tree |
| 2026-07-01 |
Lin Xiaodong | whisper : make voice_length() utf-8 aware for CJK ... |
commit | commitdiff | tree |
| 2026-06-30 |
Daniel Bevenius | ci : set GGML_NATIVE=OFF for sanitize builds (#3920) |
commit | commitdiff | tree |
| 2026-06-30 |
Yongmin Yoo 유용민 | examples : Improved inference performance of Android... |
commit | commitdiff | tree |
| 2026-06-26 |
Georgi Gerganov | talk-llama : sync llama.cpp |
commit | commitdiff | tree |
| 2026-06-26 |
Georgi Gerganov | sync : ggml |
commit | commitdiff | tree |
| 2026-06-26 |
Georgi Gerganov | ggml : bump version to 0.15.3 (ggml/1550) |
commit | commitdiff | tree |
| 2026-06-26 |
Jeff Bolz | vulkan: Workaround compiler bug in conv2d coopmat2... |
commit | commitdiff | tree |
| 2026-06-26 |
leonardHONG | CUDA: add cublasSgemmBatched mapping for HIP/MUSA vendo... |
commit | commitdiff | tree |
| 2026-06-26 |
Tarek Dakhran | ggml-cpu: fix SVE leftover path in ggml_vec_dot_f32... |
commit | commitdiff | tree |
| 2026-06-26 |
Jassieluo | sycl : clamp softmax input to avoid underflow (llama... |
commit | commitdiff | tree |
| 2026-06-26 |
leonardHONG | CUDA: batch out_prod broadcast (dps2>1) path with cubla... |
commit | commitdiff | tree |
| 2026-06-26 |
shaofeiqi | opencl: flush profiling batch at shutdown for incomplet... |
commit | commitdiff | tree |
| 2026-06-26 |
Oliver Simons | CUDA: Various fixes to `cpy.cu` (llama/25000) |
commit | commitdiff | tree |
| 2026-06-26 |
fairydreaming | ggml : address integer overflows in binary ops CUDA... |
commit | commitdiff | tree |
| 2026-06-26 |
David Spruill | sycl : support --split-mode tensor (llama/24152) |
commit | commitdiff | tree |
| 2026-06-26 |
Neo Zhang | sycl : fix the failed UT cases of conv_3d (llama/24900) |
commit | commitdiff | tree |
| 2026-06-26 |
lhez | opencl: support non-contig rows in norm (llama/24965) |
commit | commitdiff | tree |
| 2026-06-26 |
Max Krasnyansky | hexagon: MUL_MAT and MUL_MAT_ID rework : 32x32 tiled... |
commit | commitdiff | tree |
| 2026-06-26 |
Wagner Bruna | vulkan: allow reducing the graph submission batches... |
commit | commitdiff | tree |
| 2026-06-26 |
liminfei-amd | vulkan: fail the build when a shader fails to compile... |
commit | commitdiff | tree |
| 2026-06-26 |
Jeff Bolz | vulkan: Apply bias before softmax in FA, to avoid overf... |
commit | commitdiff | tree |
| 2026-06-26 |
Jeff Bolz | vulkan: support all backend tests for SQR/SQRT/SIN... |
commit | commitdiff | tree |
| 2026-06-26 |
Jeff Bolz | vulkan: Support GET_ROWS_BACK (llama/24883) |
commit | commitdiff | tree |
| 2026-06-26 |
Jeff Bolz | vulkan: support CONV_3D (llama/24612) |
commit | commitdiff | tree |
| 2026-06-26 |
Jeff Bolz | vulkan: make mul_mm ALIGNED a spec constant (llama... |
commit | commitdiff | tree |
| 2026-06-26 |
Wyatt Caldwell | vulkan: link ggml-cpu when GGML_VULKAN_CHECK_RESULTS... |
commit | commitdiff | tree |
| 2026-06-26 |
Masashi Yoshimura | ggml-webgpu: improve MTP inference by using mat-vec... |
commit | commitdiff | tree |
| 2026-06-26 |
Shawn Gu | opencl: q8_0 gemv precision improvement (llama/24923) |
commit | commitdiff | tree |
| 2026-06-26 |
Neo Zhang | support bf16 on bin_bcast OP and unary OPs (llama/24838) |
commit | commitdiff | tree |
| 2026-06-26 |
Guanhuai Zhang | fix(hexagon): use padded stride for ssm-conv weights... |
commit | commitdiff | tree |
| 2026-06-26 |
Adrien Gallouët | ggml : optimize AMX (llama/24806) |
commit | commitdiff | tree |
| 2026-06-26 |
Masashi Yoshimura | ggml-webgpu: add adapter toggles for F16 on Vulkan... |
commit | commitdiff | tree |
| 2026-06-26 |
Xuan-Son Nguyen | mtmd, arg: fix utf8 handling on windows (llama/24779) |
commit | commitdiff | tree |
| 2026-06-26 |
QuantiusBenignus | examples : fix argument flag for min speech duration... |
commit | commitdiff | tree |
| 2026-06-23 |
Daniel Bevenius | examples : update model names in parakeet-cli README... |
commit | commitdiff | tree |
| 2026-06-23 |
Nicky Mouha | Fix pkgconfig configuration (Nix build failure) (#3894) |
commit | commitdiff | tree |
| 2026-06-22 |
Naitik Shah | include parakeet in build-xcframework.sh (#3899) |
commit | commitdiff | tree |
| 2026-06-19 |
Georgi Gerganov | talk-llama : sync llama.cpp |
commit | commitdiff | tree |
| 2026-06-19 |
Georgi Gerganov | sync : ggml |
commit | commitdiff | tree |
| 2026-06-19 |
Georgi Gerganov | ggml : bump version to 0.15.2 (ggml/1548) |
commit | commitdiff | tree |
| next |
overview / pkg / ggml / sources / whisper.cpp / shortlog