| 2026-06-15 |
Jeff Bolz | vulkan: reduce iq1 shared memory usage for mul_mm ... |
commit | commitdiff | tree |
| 2026-06-15 |
Ruben Ortlam | vulkan: add `v_dot2_f32_f16` support in matrix-matrix... |
commit | commitdiff | tree |
| 2026-06-15 |
Pascal | ggml : add GGML_OP_COL2IM_1D (llama/24206) |
commit | commitdiff | tree |
| 2026-06-15 |
Yash Raj Pandey | ggml-cpu : fix rms_norm_back wrong output under in... |
commit | commitdiff | tree |
| 2026-06-15 |
ravel7524 | Remove case for GGML_TYPE_Q4_K in mvvq.cu (llama/23528) |
commit | commitdiff | tree |
| 2026-06-15 |
Reese Levine | ggml-webgpu: Add clang-format job (llama/24308) |
commit | commitdiff | tree |
| 2026-06-15 |
Masashi Yoshimura | ggml-webgpu: Improve prefill speeds for k-quants +... |
commit | commitdiff | tree |
| 2026-06-15 |
Nikhil Jain | Handle buffer overlap / buffer aliasing for concat... |
commit | commitdiff | tree |
| 2026-06-15 |
Nikhil Jain | Implement 2D workgroups for scale, binary, and unary... |
commit | commitdiff | tree |
| 2026-06-15 |
Jeff Bolz | vulkan: Use cm2 decode_vector for mul_mat_id B matrix... |
commit | commitdiff | tree |
| 2026-06-15 |
Ruben Ortlam | cuda: reset cuda context after reading memory size... |
commit | commitdiff | tree |
| 2026-06-09 |
Daniel Bevenius | ci : pin github actions to commit sha's (#3865) |
commit | commitdiff | tree |
| 2026-06-09 |
Christopher... | coreml : fix --quantize crash for mlprogram format... |
commit | commitdiff | tree |
| 2026-06-08 |
Georgi Gerganov | talk-llama : sync llama.cpp |
commit | commitdiff | tree |
| 2026-06-08 |
Georgi Gerganov | sync : ggml |
commit | commitdiff | tree |
| 2026-06-08 |
Georgi Gerganov | ggml : bump version to 0.14.0 (ggml/1533) |
commit | commitdiff | tree |
| 2026-06-08 |
Georgi Gerganov | sync : ggml |
commit | commitdiff | tree |
| 2026-06-08 |
Harkirat Gill | HIP: add gfx1152 and gfx1153 to RDNA3.5 (llama/24129) |
commit | commitdiff | tree |
| 2026-06-08 |
Xuan-Son Nguyen | metal : fix im2col 1D case (audio models) (llama/24220) |
commit | commitdiff | tree |
| 2026-06-08 |
Ruben Ortlam | vulkan: check coopmat2 features before reporting suppor... |
commit | commitdiff | tree |
| 2026-06-08 |
lhez | opencl: improve get_rows, cpy, concat and q6_k flat... |
commit | commitdiff | tree |
| 2026-06-08 |
Ruben Ortlam | vulkan: add fwht support for Intel with shmem reduction... |
commit | commitdiff | tree |
| 2026-06-08 |
Charles Xu | kleidiai : dynamic chunck-based scheduling for hybrid... |
commit | commitdiff | tree |
| 2026-06-08 |
Oliver Simons | CUDA: enroll mul_mat_vec_q_moe into pdl (llama/24087) |
commit | commitdiff | tree |
| 2026-06-08 |
Mason Milburn | sycl : port multi-column MMVQ from CUDA backend (llama... |
commit | commitdiff | tree |
| 2026-06-08 |
Kartik Sirohi | ggml: vectorize ggml_vec_dot_q4_1_q8_1 with WASM SIMD12... |
commit | commitdiff | tree |
| 2026-06-08 |
Georgi Gerganov | metal : reduce rset heartbeat from 500ms -> 5ms (llama... |
commit | commitdiff | tree |
| 2026-06-08 |
Reese Levine | ggml-webgpu: FlashAttention refactor + standardize... |
commit | commitdiff | tree |
| 2026-06-08 |
rehan-10xengineer | ggml-cpu: extend RVV quantization vec dot to higher... |
commit | commitdiff | tree |
| 2026-06-08 |
Andreas Kieslinger | Avoid PDL race conditions by disabling __restrict__... |
commit | commitdiff | tree |
| 2026-06-08 |
Charles Xu | ggml-cpu: use runtime SVE width in FWHT (llama/24059) |
commit | commitdiff | tree |
| 2026-06-08 |
Aman Gupta | cuda: reserve space for quantize kv-cache at startup... |
commit | commitdiff | tree |
| 2026-06-08 |
lhez | opencl: use flat variants of q4_K and q6_K gemv for... |
commit | commitdiff | tree |
| 2026-06-08 |
Max Krasnyansky | hexagon: profiler output fix and script updates (llama... |
commit | commitdiff | tree |
| 2026-06-08 |
Max Krasnyansky | hexagon: MUL_MAT, MUL_MAT_ID, FLASH_ATTN and GDN cleanu... |
commit | commitdiff | tree |
| 2026-06-08 |
Todor Boinovski | hexagon: add gelu_quick (llama/24007) |
commit | commitdiff | tree |
| 2026-06-08 |
Anav Prasad | clean up unused variables warnings (llama/23975) |
commit | commitdiff | tree |
| 2026-06-08 |
lhez | opencl: fix compiler warnings for non-adreno path ... |
commit | commitdiff | tree |
| 2026-06-08 |
Masashi Yoshimura | revert to using global_invocation_id for cpy shader... |
commit | commitdiff | tree |
| 2026-06-08 |
shaofeiqi | opencl: add basic support for q5_0 and q5_1 (llama... |
commit | commitdiff | tree |
| 2026-06-08 |
Shrivas Shankar | metal: template GLU kernels to support f16/f32 (llama... |
commit | commitdiff | tree |
| 2026-06-08 |
Jeff Bolz | vulkan: don't hold the device mutex while compiling... |
commit | commitdiff | tree |
| 2026-06-08 |
Winston Ma | vulkan: reduce host memory lock contention (llama/23376) |
commit | commitdiff | tree |
| 2026-06-08 |
Johannes Gäßler | TP: quantized KV cache support (llama/23792) |
commit | commitdiff | tree |
| 2026-06-08 |
Matt Corallo | vulkan: Block-load Q3_K/Q6_K block data and subtract... |
commit | commitdiff | tree |
| 2026-06-08 |
Winston Ma | vulkan: Removed unused functions (llama/23175) |
commit | commitdiff | tree |
| 2026-06-08 |
Neo Zhang | Support Q4_1, Q5_0, Q5_1 in Flash-attention (llama... |
commit | commitdiff | tree |
| 2026-06-08 |
Neo Zhang | Add more types in GET_ROWS OP (llama/23710) |
commit | commitdiff | tree |
| 2026-06-08 |
Neo Zhang | sycl : Optimize Q3_K mul_mat by reorder (llama/23725) |
commit | commitdiff | tree |
| 2026-06-08 |
lhez | opencl: support bf16 by converting to f16 (llama/23839) |
commit | commitdiff | tree |
| 2026-06-08 |
Georgi Gerganov | metal : restore im2col implementation for large kernels... |
commit | commitdiff | tree |
| 2026-06-08 |
Jinyang He | ggml : add some lsx support (llama/23798) |
commit | commitdiff | tree |
| 2026-06-08 |
Ruben Ortlam | vulkan: add Flash Attention support for BFloat16 KV... |
commit | commitdiff | tree |
| 2026-06-08 |
Reese Levine | ggml-webgpu: Check earlier for WebGPU required features... |
commit | commitdiff | tree |
| 2026-06-08 |
Reese Levine | ggml-webgpu: add q4_0/q8_0 SET_ROWS (llama/23760) |
commit | commitdiff | tree |
| 2026-06-08 |
Oliver Simons | CUDA: Check PTX version on host side to guard PDL dispa... |
commit | commitdiff | tree |
| 2026-06-08 |
fairydreaming | model : support for DeepseekV32ForCausalLM with generic... |
commit | commitdiff | tree |
| 2026-06-08 |
Daniel Bevenius | ci : add ccache to build-sycl [no ci] (#3859) |
commit | commitdiff | tree |
| 2026-06-06 |
Daniel Bevenius | ci : add HF_TOKEN to docker.yml workflow [no ci] (... |
commit | commitdiff | tree |
| 2026-06-06 |
Daniel Bevenius | ci : add ccache to quantize, vad, and wasm jobs (#3860) |
commit | commitdiff | tree |
| 2026-06-04 |
Daniel Bevenius | ci: build-windows action slimming (#3858) |
commit | commitdiff | tree |
| 2026-06-04 |
Daniel Bevenius | ci : use emscripten-core and pin version (#3857) |
commit | commitdiff | tree |
| 2026-06-04 |
Daniel Bevenius | ci : pin github actions to commit SHAs (#3856) |
commit | commitdiff | tree |
| 2026-06-04 |
Daniel Bevenius | ci : use ccache instead of sccache for windows-cublas... |
commit | commitdiff | tree |
| 2026-06-04 |
Daniel Bevenius | ci : only publish/push docker images daily (#3854) |
commit | commitdiff | tree |
| 2026-06-04 |
Georgi Gerganov | ci : refactor + optimize (#3847) |
commit | commitdiff | tree |
| 2026-06-02 |
danscMax | whisper : catch C++ exceptions in whisper_init_with_par... |
commit | commitdiff | tree |
| 2026-06-02 |
Noah Lyons | server : merge split utf-8 token text in verbose json... |
commit | commitdiff | tree |
| 2026-06-02 |
Patrice Levesque | cmake : do not assume /usr/lib library installation... |
commit | commitdiff | tree |
| 2026-06-01 |
Georgi Gerganov | release : v1.8.6 |
commit | commitdiff | tree |
| 2026-06-01 |
Daniel Bevenius | ci : fix path to whisper.h in examples.yml [no ci]... |
commit | commitdiff | tree |
| 2026-05-31 |
Georgi Gerganov | ci : fix self-hosted paths to mnt |
commit | commitdiff | tree |
| 2026-05-31 |
Georgi Gerganov | pi : add config |
commit | commitdiff | tree |
| 2026-05-31 |
Georgi Gerganov | ci : remove obsolete self-hosted label |
commit | commitdiff | tree |
| 2026-05-31 |
Georgi Gerganov | common : pass sample rate to `ffmpeg_decode_audio()` |
commit | commitdiff | tree |
| 2026-05-31 |
Georgi Gerganov | common : re-implement `ffmpeg-transcode.cpp` + clarify... |
commit | commitdiff | tree |
| 2026-05-29 |
Georgi Gerganov | sync : ggml |
commit | commitdiff | tree |
| 2026-05-29 |
Georgi Gerganov | ggml : bump version to 0.13.1 (ggml/1523) |
commit | commitdiff | tree |
| 2026-05-29 |
Georgi Gerganov | talk-llama : sync llama.cpp |
commit | commitdiff | tree |
| 2026-05-29 |
Georgi Gerganov | sync : ggml |
commit | commitdiff | tree |
| 2026-05-29 |
Andreas Kieslinger | cuda : disables launch_fattn PDL enrollment due to... |
commit | commitdiff | tree |
| 2026-05-29 |
Matt Corallo | meta : Add missing `buffer` set in allreduce fallback... |
commit | commitdiff | tree |
| 2026-05-29 |
Max Krasnyansky | hexagon: basic/generic op fusion support and RMS_NORM... |
commit | commitdiff | tree |
| 2026-05-29 |
lhez | opencl: move backend info printing into its own functio... |
commit | commitdiff | tree |
| 2026-05-29 |
fl0rianr | ggml: auto apply iGPU flag CUDA/HIP if integrated devic... |
commit | commitdiff | tree |
| 2026-05-29 |
redfox | mmvq Optim: add MMVQ_PARAMETERS_TURING(mmvq_parameter_t... |
commit | commitdiff | tree |
| 2026-05-29 |
Jaden_Mach | CUDA: route batch>=4 quantized matmul to MMQ on AMD... |
commit | commitdiff | tree |
| 2026-05-29 |
Max Krasnyansky | hexagon: minor refresh for HMX FA and MM (llama/23796) |
commit | commitdiff | tree |
| 2026-05-29 |
Jeff Bolz | vulkan: fast path for walsh-hadamard transform (llama... |
commit | commitdiff | tree |
| 2026-05-29 |
Winston Ma | vulkan: fix wrong index variable in inner loop (llama... |
commit | commitdiff | tree |
| 2026-05-29 |
Winston Ma | vulkan: Fix memory logger unsafe iterator access (llama... |
commit | commitdiff | tree |
| 2026-05-29 |
fairydreaming | cuda : fix KQ mask offset integer overflow in fattn... |
commit | commitdiff | tree |
| 2026-05-29 |
Martin Klacer | ggml: fixed Arm SVE usage bug in vec.h, vec.cpp (llama... |
commit | commitdiff | tree |
| 2026-05-29 |
ymcki | Hexagon: OP_GATED_DELTA_NET K>1 support (llama/23531) |
commit | commitdiff | tree |
| 2026-05-29 |
ymcki | opencl: OP_GATED_DELTA_NET (llama/23312) |
commit | commitdiff | tree |
| 2026-05-29 |
Reese Levine | ggml-webgpu: remove legacy constants (llama/23672) |
commit | commitdiff | tree |
| 2026-05-29 |
Max Krasnyansky | hexagon: add support for Q4_1 in MUL_MAT and MUL_MAT_ID... |
commit | commitdiff | tree |
| 2026-05-29 |
Masashi Yoshimura | ggml-webgpu: Fix how to dispatch WG to some ops (llama... |
commit | commitdiff | tree |
| 2026-05-29 |
Matt Corallo | vulkan: Switch MUL_MAT_VEC to 4 K per iteration for... |
commit | commitdiff | tree |
| 2026-05-29 |
Jeff Bolz | vulkan: use GL_NV_cooperative_matrix_decode_vector... |
commit | commitdiff | tree |
| next |
overview / pkg / ggml / sources / whisper.cpp / shortlog