ggml-org/llama.cpp

Pull Requests Commits

cli: add --output option

ngxson committed 29 days ago

7ab9323d

llama-bench : init params.offline (#25476)

angt committed 30 days ago

Verified 259f2e2a

metal : add CONV_2D_DW (depthwise convolution) support (#21565)

Sou-ly committed 30 days ago

Verified 92b187c9

ggml-hip: enable -funsafe-math-optimizations (#24668)

RapidMark committed 30 days ago

Verified ccb0c342

cuda: align snake fusion matcher with the other backends (#25460)

ServeurpersoCom committed 30 days ago

Verified 2021515a

server : respect min-step when splitting prompt batches (#25420)

aldehir committed 30 days ago

Verified 64c8b7db

hexagon: add VISION RoPE support (#25216)

aparmp-quic committed 30 days ago

Verified f2d1c2f3

ggml-webgpu: tune subgroup split (d_split) in flash_attn_vec (#25418)

yomaytk committed 30 days ago

Verified 32e41fa5

opencl: Q6_K GEMM/GEMV fix for ne01 of weights that are not multiples of 128. (#25464)

wanghqc committed 30 days ago

Verified 92366df3

vulkan: disable FA mask_opt on GCN to improve performance (#24362)

0cc4m committed 30 days ago

Verified a646006f

opencl: ragged-tile MoE prefill FP16 GEMM optimization (skip padded expert tiles) (#25433)

wanghqc committed 30 days ago

Verified 167d0576

llama-batch: fix allowed decreasing pos in a seq (#25449)

am17an committed 30 days ago

Verified 1ee09393

vulkan: for small AMD GPUs, reduce submission threshold based on CU count (#25240)

0cc4m committed 30 days ago

Verified 0bbc87b1

hexagon: new vtcm layouts and improved pipelines for MUL_MAT, MUL_MAT_ID and FLASH_ATTN_EXT (#25425)

max-krasnyansky committed 30 days ago

Verified 81ff7abe

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

ngxson committed 30 days ago

Verified c264f65f

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

ORippler committed 30 days ago

Verified 07e012af

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

fairydreaming committed 30 days ago

Verified ed8c2615

llama: refactor fused ops (#24646)

am17an committed 31 days ago

Verified 90e0f5cf

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

ServeurpersoCom committed 31 days ago

Verified bbebeec4

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

am17an committed 31 days ago

Verified 230ea9d2

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

rankaiyx committed 31 days ago

Verified f296fdfb

ui: Context usage gauge and panel (#25340)

allozaur committed 31 days ago

Verified f1161b15

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

ggerganov committed 31 days ago

Verified da46e59c

metal : add set_rows with src0 f16 (#25434)

fairydreaming committed 31 days ago

Verified 0512ef1e

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

hourhl committed 31 days ago

Verified 4a7ee312

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

tyronecai committed 31 days ago

Verified 57b50e1f

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

fairydreaming committed 31 days ago

Verified 68a521b5

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

lhez committed 31 days ago

Verified 931ca30b

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

khosravipasha committed 31 days ago

Verified bec4772f

spec : fix naming, spacing (#25410)

ggerganov committed 31 days ago

Verified c198af4d

Older