ggml-org/llama.cpp

Pull Requests Commits

remove unnecessary host write pipeline barriers

0cc4m committed 12 hours ago

0a1ebc5d

vulkan: add pipeline barriers for memcpy read/write operations

0cc4m committed 19 hours ago

156c85bb

ggml-zendnn : fixed naming of matmul function (#20964)

truecoder34 committed 1 day ago

Verified b4c0549a

ci : do not allocate ccache for 3rd-party hosted runners (#23730)

ggerganov committed 1 day ago

Verified 0d18aaa9

ci : move [no release] check to dedicated check_release job (#23734)

ggerganov committed 1 day ago

Verified 08bc21b4

ci : add `[no release]` keyword + fix sanitizer builds (#23728)

ggerganov committed 1 day ago

Verified 35a74c8f

ci : move macos jobs to the apple workflow + fix names (#23721)

ggerganov committed 1 day ago

Verified 5190c2ea

vulkan: optimize conv2d and implement coopmat1 support (#22620)

jeffbolznv committed 1 day ago

Verified 7799d31e

ci : remove vulkan SDK dep from webgpu job (#23718)

ggerganov committed 1 day ago

Verified 3a3ed153

hexagon: add support for CONCAT op (#23648)

max-krasnyansky committed 1 day ago

Verified ef66bfab

ci : move more CPU jobs to self-hosted runners (#23715)

ggerganov committed 1 day ago

Verified 678d43d7

ci : move sanitizer jobs to self-hosted runners (#23713)

ggerganov committed 1 day ago

Verified ef41a691

ci : reduce (disable SYCL and CANN builds/releases) (#23705)

ggerganov committed 1 day ago

Verified 3dc7684f

convert : support Gemma4ForCausalLM architecture (#23682)

aoleg committed 1 day ago

Verified dbe9c0c8

models : Attach Mistral3 NVFP4 weight scales (#23629)

michaelw9999 committed 1 day ago

Verified 6fe90def

SYCL: implement ggml_sycl_pool_vmm (#22862)

sanmai committed 1 day ago

Verified 581d020b

tests: test-backend-ops -j <N> to run tests in parallel (#23637)

jeffbolznv committed 1 day ago

Verified 7623de11

model : add support for talkie-1930-13b (#22596)

niklassheth committed 1 day ago

Verified c9d98295

ggml-webgpu: Add MMVQ path for Q4/Q8/Q2_K/Q4_K and clean up legacy MUL_MAT pipeline (#23594)

yomaytk committed 2 days ago

Verified 1506d39e

[WebGPU] Check batch_compute_passes before sending passes when not doing GPU profiling (#23457)

nikhilJain17 committed 2 days ago

Verified 54121f73

CUDA: missing PDL sync for FWHT, better fallback (#23690)

JohannesGaessler committed 2 days ago

Verified 192d8ae8

metal : add apple device id (#23566)

forforever73 committed 2 days ago

Verified 35c9b1f3

snapdragon: bump toolchain docker to v0.7 to fix ui build issues (#23680)

max-krasnyansky committed 2 days ago

Verified 4bead4e3

ci : reduce PR jobs by matching backend paths (#23675)

ggerganov committed 2 days ago

Verified 302e2c26

model: tag ffn_latent as MUL_MAT to fix buft probe (#23664)

ServeurpersoCom committed 2 days ago

Verified 328874d0

CUDA: add fast walsh-hadamard transform (#23615)

am17an committed 2 days ago

Verified c1f1e28d

ui: fix stop/continue during an agentic loop (#23356)

ServeurpersoCom committed 2 days ago

Verified 5a4126ad

convert : add compressed-tensors NVFP4 support (#21095)

michaelw9999 committed 2 days ago

Verified a4d2d4ae

ggerganov committed 2 days ago

d161ea70

ggml : bump version to 0.13.0 (ggml/1510)

ggerganov committed 2 days ago

45158f46

Older