ggml-org/llama.cpp

Pull Requests Commits

label virtual devices in description; add GPUx2 server CI jobs

anavp-nvidia committed 23 days ago

234252e0

disable NCCL path when virtual devices are used

anavp-nvidia committed 29 days ago

2976e33c

support cuda virtual devices

anavp-nvidia committed 29 days ago

a8eca04c

cli : move to HTTP-based implementation (#24948)

ngxson committed 29 days ago

Verified c264f65f

Make hip quality check run on all changes (#25403)

ORippler committed 29 days ago

Verified 07e012af

cuda : add support for f16->f16 GGML_OP_SET_ROWS (#25367)

fairydreaming committed 29 days ago

Verified ed8c2615

llama: refactor fused ops (#24646)

am17an committed 29 days ago

Verified 90e0f5cf

server-stream: follow-up on SSE Replay Buffer (#23226) (#25047)

ServeurpersoCom committed 29 days ago

Verified bbebeec4

llama-batch: add n_keep_tail in split_equal for recurrent models (#25278)

am17an committed 29 days ago

Verified 230ea9d2

common: auto-create prompts-log-dir at argument parsing, so all tools using the flag benefit (#25322)

rankaiyx committed 29 days ago

Verified f296fdfb

ui: Context usage gauge and panel (#25340)

allozaur committed 29 days ago

Verified f1161b15

llama-eval : fix crash when answer is None in HTML dump (#25435)

ggerganov committed 29 days ago

Verified da46e59c

metal : add set_rows with src0 f16 (#25434)

fairydreaming committed 29 days ago

Verified 0512ef1e

fix: OOB reads in UGM tokenizer (precompiled_charsmap handling) (#18750)

hourhl committed 29 days ago

Verified 4a7ee312

ggml : fix A indexing in simd_gemm scalar tail-column path (#25390)

tyronecai committed 29 days ago

Verified 57b50e1f

ggml : add support for CPU f16->f16 GGML_OP_SET_ROWS (#25344)

fairydreaming committed 29 days ago

Verified 68a521b5

opencl: fix potential crash in aos reconstruct (#25383)

lhez committed 29 days ago

Verified 931ca30b

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

khosravipasha committed 29 days ago

Verified bec4772f

spec : fix naming, spacing (#25410)

ggerganov committed 30 days ago

Verified c198af4d

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

ORippler committed 30 days ago

Verified 3899b39c

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

wadealexc committed 30 days ago

Verified f5525f7e

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

thom-dev-fr committed 30 days ago

Verified 5eca4e3c

server: enforce prompt cache RAM limit (#25070)

tarruda committed 30 days ago

Verified 6c487e2f

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

zhangrunda committed 30 days ago

Verified c1a411fb

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

asf0 committed 30 days ago

Verified 33ca0dcb

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

am17an committed 30 days ago

Verified 024c46ae

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

arthw committed 30 days ago

Verified 108f186d

[SYCL] support op col2im_1d (#25264)

arthw committed 30 days ago

Verified 47e1de77

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

arthw committed 30 days ago

Verified 55edb2de

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

malsbat committed 30 days ago

Verified d2090861

Older