make memset range dynamic#13002

Closed

pockers21 wants to merge 54 commits intoggml-org:masterfrom

pockers21:feature-dynamic-memset-range

Commits on Apr 18, 2025

make memset range dynamic
pockers21
committed

Commits on Apr 28, 2025

rpc : add RPC_CMD_HELLO (#12955)

rgerganov
authored and
pockers21
committed
mtmd : add methods to access `mtmd_image_tokens` (#12906)

ngxson
authored and
pockers21
committed
SYCL: Refactor and enable FP16 in binary broadcast OPs (#12975)

qnixsynapse
authored and
pockers21
committed
server : use std::move whenever possible (#12936)

authored and
pockers21
committed
gguf-py : GGUF Editor GUI - Python + Qt6 (#12930)

christopherthompson81
authored and
pockers21
committed
main : Fix Ctrl+D/newline handling (#12951)

danielzgtg
authored and
pockers21
committed
clip : refactor, add `image_manipulation` and `llava_uhd` classes (#13011)

ngxson
authored and
pockers21
committed
gguf-py : fix upload python package workflow (#13020)

CISC
authored and
pockers21
committed
Disable CI cross-compile builds (#13022)

bandoti
authored and
pockers21
committed
metal: add neg operator (#13029)

jmorganca
authored and
pockers21
committed
vulkan: support noncontiguous rms_norm (#13031)

jeffbolznv
authored and
pockers21
committed
llava: fix errors in clip.h on certain compilers (#13030)

jmorganca
authored and
pockers21
committed
convert : experimental support for `--mmproj` flag (#13023)

authored and
pockers21
committed
mtmd : merge llava, gemma3 and minicpmv CLI into single `llama-mtmd-cli` (#13012)

authored and
pockers21
committed
SYCL: Add non-contiguous support in ROPE (#12993)

qnixsynapse
authored and
pockers21
committed
ggml : add SSE 4.2 and x64 base variant for CPUs without AVX (#12871)

slaren
authored and
pockers21
committed
llava : update documentations (#13055)

ngxson
authored and
pockers21
committed
metal : add memory pool for temp allocs (#12850)

ggerganov
authored and
pockers21
committed
security : add note about RPC and server functionality (#13061)

ggerganov
authored and
pockers21
committed
mtmd : support SmolVLM (version 1 and 2) (#13050)

ngxson
authored and
pockers21
committed
CUDA: noncont MMVQ + batched bs1 MUL_MAT_ID (#13014)

JohannesGaessler
authored and
pockers21
committed
rpc : add command line option for number of threads for the CPU backend (#13060)

rgerganov
authored and
pockers21
committed
convert : Append mult-eos,half-rope,bos to GLM4-0414 and Z (#13021)

piDack
authored and
pockers21
committed
mtmd : Support Pixtral 12B (#13065)

ngxson
authored and
pockers21
committed
llama-mtmd-cli: Sigint rework in mtmd vision example (#13080)

authored and
pockers21
committed
vulkan: matmul gcn tuning (#13016)

authored and
pockers21
committed
metal : fix floating-point range of attention scores in FA kernels (#13090)

ggerganov
authored and
pockers21
committed
arg : clean up handling --mmproj with -hf (#13082)

ngxson
authored and
pockers21
committed
arg : add --no-mmproj-offload (#13093)

ngxson
authored and
pockers21
committed
clang-tidy : disable warning about missing math parenthesis (#13091)

ggerganov
authored and
pockers21
committed
cmake : do not include ./src as public for libllama (#13062)

ggerganov
authored and
pockers21
committed
CUDA: use switch statements in constexpr functions (#13095)

JohannesGaessler
authored and
pockers21
committed
ggml : Depthwise 2D convolution (ggml/1152)

Acly
authored and
pockers21
committed
sync : ggml

ggerganov
authored and
pockers21
committed
ggml : fix trailing whitespaces (#0)

ggerganov
authored and
pockers21
committed
embeddings : fix batch sizes (#13076)

ggerganov
authored and
pockers21
committed
clip : remove boi/eoi embeddings for GLM-edge model (#13081)

ngxson
authored and
pockers21
committed
rpc : do not wait for response when sending RPC_CMD_SET_TENSOR (#12943)

rgerganov
authored and
pockers21
committed
change the reorder tensor from init to execute OP (#13003)

NeoZhangJianyu
authored and
pockers21
committed
clip : fix pixtral on some GPU backends (#13097)

ngxson
authored and
pockers21
committed
Force FP32 compute in GLM4 FFN Down (#13101)

authored and
pockers21
committed
llama : fix K-shift with quantized K and BLAS backend (#13113)

slaren
authored and
pockers21
committed
grammar : handle maxItems == 0 in JSON schema (#13117)

authored and
pockers21
committed
ggml: move fp16/bf16 conversion optimizations to CPU backend + export conversion APIs (#13107)

SongXiaoXi
authored and
pockers21
committed
clip : improve projector naming (#13118)

ngxson
authored and
pockers21
committed
common : add common_remote_get_content (#13123)

ngxson
authored and
pockers21
committed
clip : Add Qwen2.5VL support (#12402)

authored and
pockers21
committed
Fixes Qwen2.5VL segfault during inference with https://github.com/ggml-org/llama.cpp/pull/12402 as has_qwen2vl_merger migration was incomplete (#13133)

LostRuins
authored and
pockers21
committed
musa: fix build warning (#13129)

yeahdongcn
authored and
pockers21
committed
llama-chat : fix wrong template in GLM4-0414 (#13140)

matteoserva
authored and
pockers21
committed
llama-bench : Add `--override-tensors` arg (#12922)

4onen
authored and
pockers21
committed
arg : fix unused variable (#13142)

ngxson
authored and
pockers21
committed
delete set size and delete buffer clear as suggested
pockers21
committed