ngxson/llama.cpp

Pull Requests Commits

Add Q2_0 quantization: type definition and CPU backend (#24448)

khosravipasha committed 78 days ago

Verified bec4772f

spec : fix naming, spacing (#25410)

ggerganov committed 78 days ago

Verified c198af4d

CUDA: Fuse MMVQ post-scale for NVFP4 (#24481)

ORippler committed 78 days ago

Verified 3899b39c

server : fix draft model fit vs load inconsistency (#25056)

wadealexc committed 78 days ago

Verified f5525f7e

server : add timings and progress to /responses API stream (#25348)

thom-dev-fr committed 78 days ago

Verified 5eca4e3c

server: enforce prompt cache RAM limit (#25070)

tarruda committed 78 days ago

Verified 6c487e2f

common : add missing <fstream> include in common.h (#25220)

zhangrunda committed 78 days ago

Verified c1a411fb

ggml-hip : add -fno-finite-math-only alongside -ffast-math (#25373)

asf0 committed 78 days ago

Verified 33ca0dcb

llama: fix quantized kv-cache for dsv4 (#25202)

am17an committed 78 days ago

Verified 024c46ae

[SYCL] fix unsupported UT cases of CONT & CPY (#25231)

arthw committed 78 days ago

Verified 108f186d

[SYCL] support op col2im_1d (#25264)

arthw committed 78 days ago

Verified 47e1de77

[SYCL] support OP cross_entropy_loss, cross_entropy_loss_back (#25236)

arthw committed 78 days ago

Verified 55edb2de

sycl : set K_QUANTS_PER_ITERATION to 1 on DMMV path (#25063)

malsbat committed 78 days ago

Verified d2090861

[SYCL] fix unsupport ACC UT cases for noncontiguous (#25124)

arthw committed 78 days ago

Verified 95e5254c

sycl : enhance argsort to support all UT cases (#25125)

arthw committed 78 days ago

Verified 9e5ef0db

sycl : use sycl func to fix AOT double type issue (#25081)

arthw committed 78 days ago

Verified 3d4cbdf1

sycl : rename the env vars from "disable" to "enable" (#25042)

arthw committed 78 days ago

Verified 26145b3d

ggml : make ggml_time_init idempotent (#24422)

aisk committed 78 days ago

Verified 1a7c25bf

speculative : fix out-of-bounds read in ngram-map on prompt shrink (#23936)

o7si committed 78 days ago

Verified defa95c3

vulkan : check src0 type in GGML_OP_SET_ROWS to avoid failures due to unimplemented f16 support (#25351)

fairydreaming committed 78 days ago

Verified a8cfdbb9

opencl: general flash attention decode performance optimizations (#25366)

wanghqc committed 79 days ago

Verified 6f8895fe

common: Set optimal default thread count for ppc ( linux as well as AIX) (#25237)

shalinib-ibm committed 79 days ago

Verified ee445f93

metal: add col2im_1d op (f32/f16/bf16) (#25176)

ServeurpersoCom committed 79 days ago

Verified f36e5c34

CUDA: remove -sm row, refactor cuBLAS (#24216)

JohannesGaessler committed 79 days ago

Verified 74976e1a

server: fix deadlock in load_models() when erasing a finished download (#25358)

ServeurpersoCom committed 79 days ago

Verified 9abce747

CUDA: extend K-type validation to V-types for flash attention (#24403)

sanmai committed 79 days ago

Verified cb295bf5

server: temporary skip model downloading API test (#25355)

ngxson committed 79 days ago

Verified bfdf581b

ggml-cpu: use UE4M3 LUT in ARM NVFP4 dot product (#25331)

ragz4125 committed 79 days ago

Verified 20a04b22

ggml-cpu: Enable tiled matmul on AIX (#25199)

shalinib-ibm committed 79 days ago

Verified 3b4fca11

vulkan: fix 32-bit integer overflow in CEIL_DIV (#25245)

hokanosekai committed 79 days ago

Verified 86961efd

Newer Older