ggml-org/llama.cpp

Pull Requests Commits

quantize: row-slab stream to avoid thread starvation

ngxson committed 22 days ago

f47ff9b2

server: add ctx-per-slot (--kv-unified-per-slot) (#24124)

bartowski1182 committed 22 days ago

Verified 18443257

ci : build only the ggml-hip backend for windows-rocm release (#27753)

harkgill-amd committed 22 days ago

Verified 32176338

model: add Qwen3.8-Flash-Next (qwen4exp) (#27742)

danielhanchen committed 22 days ago

Verified 6c84c7d5

ci : bundle HIP runtime DLLs with Windows ROCm release (#26973)

slojosic-amd committed 22 days ago

Verified 6fdd0ac8

spec : add DFlash2 support (local convolution + candidate selector) (#27342) (#27816)

ngxson committed 22 days ago

Verified b10f9ca5

opencl: add bin kernels `kernel_gemm_moe_q4_0_q8_1_dp4a_bin`, `kernel_gemm_moe_mxfp4_q8_1_dp4a_bin` (#27768)

shawngu-quic committed 22 days ago

Verified 58546250

quantize: cap working memory size to avoid loading big tensors onto RAM (#27795)

ngxson committed 22 days ago

Verified 732707df

Feature: Added LIGHTNING_INDEXER support for Deepseek V4 ops on Vulkan Backend (#27453)

shenron0101 committed 23 days ago

Verified cb300598

pr2wt : use ssh/https remote in worktree depending on base (#27800)

CISC committed 23 days ago

Verified 1a946ec7

llama: model_loader: add TENSOR_READ_LAZY (#27794)

ngxson committed 23 days ago

Verified fac889fb

ui: Improve Chat Form Actions UI/UX (models selector, add panel) (#27746)

allozaur committed 23 days ago

Verified cae63579

convert : fix Nemotron-H LoRA GGUF conversion (#27356)

frozenblade1224 committed 23 days ago

Verified bcb6084a

ui: Replace per-conversation MCP overrides with per-conversation tool policy (#27745)

allozaur committed 23 days ago

Verified fe235f43

spec: Add benchmark-only synthetic speculative acceptance options (#27711)

gaugarg-nv committed 23 days ago

Verified 2bb9bdda

model : simplify MiniMax-01 graph (#27790)

fairydreaming committed 23 days ago

Verified deae5ee1

args: add --video-* CLI arguments (#24318)

ngxson committed 23 days ago

Verified f2955121

metal : fix memory leaks due to missing autoreleasepools (#27758)

nikwen committed 23 days ago

Verified 915dc6d3

llama : add --n-cpu-ffn option (#26622)

John-194 committed 23 days ago

Verified c5fc7e34

models : support nanbeige4.2-3B (#27730)

zqlcode committed 23 days ago

Verified d7a20741

hexagon: support for multi-NPU devices (IQ9, IQ10) and fully asynchronous backend (#26501)

max-krasnyansky committed 23 days ago

Verified 192067b7

llama: add token ID tracking to KV cell (#27762)

ngxson committed 23 days ago

Verified 925e1179

ui: Move Settings and MCP Servers routes to dialog-based views (#27744)

allozaur committed 23 days ago

Verified 539f2452

ui: Update Dialog component styling (#27743)

allozaur committed 23 days ago

Verified 0379a19f

vulkan: warptiles currently assume warp sizes <= 64, clamp to work around larger warps (#27726)

0cc4m committed 23 days ago

Verified 5e6a37cb

Implemented vulkan cross_entropy_loss and cross_entropy_loss_back (#27216)

PranavUttarkar committed 24 days ago

Verified bf942164

rpc : implement event and async backend APIs (#18626)

rgerganov committed 24 days ago

Verified d0132a68

ci: Clean up UI builds from releases (#27706)

allozaur committed 24 days ago

Verified 4d19b287

cuda: unblock mmq for MoE on sm_60 (#26264)

dfriehs committed 24 days ago

Verified fc35562b

ci : make cache bucket public (#27728)

CISC committed 24 days ago

Verified da9b5d68

Older