ggml-org/llama.cpp

Pull Requests Commits

ngxson committed 23 days ago

f4bd37bd

ngxson committed 23 days ago

5676c4bd

only allow is_first_load to populate it

ngxson committed 23 days ago

166cad8a

server: (router) lazy-load startup_models after main setup

ngxson committed 24 days ago

c47322cc

tensor-split meta backend fixes (#26502)

max-krasnyansky committed 24 days ago

Verified d59d455f

hexagon: fix FA HMX queue ordering and pack the rescale D matrices (#27042)

njsyw1997 committed 24 days ago

Verified 990e3bfe

opencl: port fused ssm_scan kernel (Mamba-2, d_state in {128, 256}) to GPU (#26439)

wanghqc committed 24 days ago

Verified b062ba73

ggml-cpu: gate __fp16 on __ARM_FP16_FORMAT_IEEE (#26860)

ServeurpersoCom committed 24 days ago

Verified cd644c39

server: refactor sleep handling, allow access /metrics during sleep (#27376)

ngxson committed 24 days ago

Verified 947fd9bb

server : make models endpoints private when authentication is enabled (#26347)

s0mecode committed 24 days ago

Verified ee0ea03a

vulkan : dequant q8_0 KV once in coopmat1 (#25494)

Nathanw1014 committed 24 days ago

Verified dc72703f

vulkan: add null checks in ggml_vk_queue_command_pools_cleanup (#27353)

jetsontan committed 24 days ago

Verified b95502ba

Revert "common: share thread pools when `n_threads` differ (#27138)" (#27337)

nikwen committed 24 days ago

Verified 3e734467

model : GraniteSWAForCausalLM / GraniteMoeSWAForCausalLM (#25505)

gabe-l-hart committed 24 days ago

Verified 7221e24f

sycl: report zero devices instead of aborting when the host has none (#27291)

Titaniumtown committed 24 days ago

Verified 6cc504a2

ci : add release attestation url (#27389)

CISC committed 24 days ago

Verified 01ac3ad7

models : remove duplicate metadata load (#27378)

CISC committed 24 days ago

Verified 2e92ecd0

ci : re-enable release dependency for sycl (#27385)

CISC committed 24 days ago

Verified 645ca283

ggml: add ggml_rope_set_offset (+ metal support) (#27120)

ngxson committed 24 days ago

Verified fe8156f7

ui: read persisted settings before the API key probe (#27365)

ServeurpersoCom committed 24 days ago

Verified 77acca43

mtmd: add mtmd_bitmap_set_mergeable (#27348)

ngxson committed 24 days ago

Verified 95c409c1

metal : dequantize q8_0 using packed types (#27370)

ggerganov committed 24 days ago

Verified 8ef78e64

server: add dedup-cache-models preset option (#27346)

ngxson committed 24 days ago

Verified ee4c505a

vulkan: tiled transpose for 0<->2 permuted CONT (#26585)

kh0pper committed 24 days ago

Verified 98d1e92c

ggml-webgpu: add mulmat with overlapping src0/src1 (e.g., for minimax-01) (#27321)

yomaytk committed 24 days ago

Verified 5112b973

ci : add attestation for signed release artifacts (#25933)

quartzjer committed 24 days ago

Verified 0adcc3bb

tests: skip test-unicode build on win32/BUILD_SHARED_LIBS (#27336)

jeffbolznv committed 24 days ago

Verified 79fe7991

gguf-py : add size guards to GGUFReader (#27188)

oakkaya committed 24 days ago

Verified 0329fcda

server: (cosmetic) do not print cmd_child_to_router messages [no release] (#27347)

ngxson committed 25 days ago

Verified 6d054983

opencl: fix WAR race in the generic FA tile kernels when the WG spans subgroups (#26434)

wanghqc committed 25 days ago

Verified 60addddf

Older