ngxson/llama.cpp

Pull Requests Commits

webui: introduce OpenAI-compatible model selector in JSON payload (#16562)

ServeurpersoCom committed 157 days ago

Verified 9b9201f6

ggml : Leverage the existing GGML_F32_VEC helpers to vectorize ggml_vec_set_f32 for faster fills (#16522)

sirus20x6 committed 157 days ago

Verified 19a5a3ed

tests : fix test-thread-safety when compiling with multiple backends (#16699)

Acly committed 157 days ago

Verified d8eaa26e

CUDA: fix bug in topk-moe softmax (#16711)

am17an committed 157 days ago

Verified 9285325c

CUDA: topk-moe: add optional parameter for gpt-oss (#16649)

am17an committed 158 days ago

Verified 03792ad9

CUDA: better error for FA kernel with 0 occupancy (#16643)

JohannesGaessler committed 158 days ago

Verified 51d1a8c9

ggml: add ggml_can_fuse_subgraph (#16662)

am17an committed 158 days ago

Verified 4926419c

opencl: fix warnings and clean up profiling (#16688)

lhez committed 158 days ago

Verified 6ea37f57

vulkan: Handle FA with all -inf mask values (#16447)

jeffbolznv committed 158 days ago

Verified fb349848

sycl : add PAD_REFLECT_D1 operator support (#16145)

ye-NX committed 159 days ago

Verified 6de8ed75

model : add BailingMoeV2 support (#16063)

CISC committed 159 days ago

Verified 84bf3c67

Handle legacy 'context' attachments (#16687)

allozaur committed 159 days ago

Verified c9c1972e

ggml-alloc : fix leak when reusing a tensor with a larger size (#16679)

slaren committed 159 days ago

Verified b617cfd2

Prevent premature submission on IME input (#16673)

allozaur committed 159 days ago

Verified 79068501

Import/Export UX improvements (#16619)

allozaur committed 159 days ago

Verified 0e4a0cf2

Enable per-conversation loading states to allow having parallel conversations (#16327)

allozaur committed 159 days ago

Verified 13f2cfad

llama-batch: fix build fails with `-Werror=missing-braces` (#16614)

otegami committed 159 days ago

Verified 06332e28

readme: update bindings (#16651)

deadprogram committed 159 days ago

Verified 72d53e6c

SYCL: Add support for FLOOR,CEIL,ROUND and TRUNC unary operators (#16613)

safranowith committed 159 days ago

Verified 2330de7b

llama-context: only warn on pooling_type when user specified (#16674)

otegami committed 159 days ago

Verified 7062dd84

model : add Granite Hybrid types (#16635)

giuseppe committed 160 days ago

Verified 0398752d

ci : fix binaries release failure for s390x (binaries may not work yet) (#16664)

taronaeo committed 160 days ago

Verified 4f73d0a9

ci : avoid manual updates of docs/ops.md (#16663)

CISC committed 160 days ago

Verified cec5edbc

ci: include s390x release binaries (#16648)

taronaeo committed 160 days ago

Verified fcb235b4

CODEOWNERS: update for ggml-cuda/mmf (#16660)

am17an committed 160 days ago

Verified 55754beb

HIP: fix GPU_TARGETS (#16642)

JohannesGaessler committed 161 days ago

Verified ee09828c

vulkan: Implement topk_moe fused shader, ported from CUDA (#16641)

jeffbolznv committed 161 days ago

Verified e56abd20

CUDA: use registers instead of smem in topk-moe (#16647)

am17an committed 161 days ago

Verified 38355c6c

opencl: transposed gemm/gemv moe kernel with mxfp4,f32 (#16602)

shawngu-quic committed 162 days ago

Verified 81387858

llama-model: fix insonsistent ctxs <-> bufs order (#16581)

JohannesGaessler committed 162 days ago

Verified 66b0dbcb

Newer Older