Merge upstream llama.cpp updates into spacemit-mtmd#6

Open

co-seven wants to merge 310 commits into

spacemit-com:spacemit-mtmdfrom

co-seven:merge-upstream-20260602

This pull request is big! We're only showing the most recent 250 commits

Commits on May 18, 2026

Commits on May 19, 2026

Commits on May 20, 2026

Commits on May 21, 2026

app : show version (#23426)
angt
authored
hexagon: ssm-conv fix for large prompts (#23307)

tboinovski1
and
max-krasnyansky
authored
llama-graph: fix null-buffer crash in llm_graph_input_attn_kv_iswa for SWA-only models (#23131)
ssfdre38
authored
ggml : Check the right iface method before using the fallback 2d get (#23306)
TheBlueMatt
authored
ui: Improve Git Hooks for UI development (#23403)

allozaur
and
ServeurpersoCom
authored
doc: fix spec mtp typo (#23435)
ruixiang63
authored
vocab : add Carbon-3B (HybridDNATokenizer) support (#23410)

kashif
and
CISC
authored
mtp: use inp_out_ids for skipping logit computation (#23433)
am17an
authored
app : add batched-bench, fit-params, quantize & perplexity (#23459)
angt
authored
server: re-inject subcommand when router spawns children under unified binary (#23442)
ServeurpersoCom
authored
server : free draft/MTP resources on sleep to fix VRAM leak (#23461)
am17an
authored
metal : optimize concat kernel and fix set kernel threads (#23411)
ggerganov
authored
server: expose prompt token counts in /slots endpoint (#23454)
ScrewTSW
authored
tests : move save-load-state from examples to tests (#23336)
ggerganov
authored
fix(flash-attn): replace f32 with kv_type and q_type (#23372)
Constannnnnt
authored
vulkan: fuse snake activation (mul, sin, sqr, mul, add) (#22855)
ServeurpersoCom
authored
Update WebGPU support and add link to blog/demo (#23483)
reeselevine
authored
cmake : remove STATIC from impl libraries, enable LLAMA_BUILD_APP by default (#23462)
ggerganov
authored
CUDA: fix PDL CC check for JIT compilation (#23471)
JohannesGaessler
authored

Commits on May 22, 2026

Commits on May 24, 2026

Commits on May 25, 2026

Commits on May 26, 2026

Commits on May 27, 2026

Commits on May 28, 2026

Commits on May 29, 2026

Commits on May 30, 2026

Commits on Jun 1, 2026

Commits on Jun 3, 2026

fix server converter warning errors
co-seven
committed