Sync upstream master into hybrid cache work branch - #1

Merged

Regrad merged 134 commits into

fix/qwen-hybrid-cache-radicalfrom

sync/upstream-master-2026-06-07

Jun 7, 2026

Commits on May 29, 2026

Commits on May 30, 2026

Commits on Jun 1, 2026

Commits on Jun 2, 2026

opencl: fix compiler warnings for non-adreno path (#23922)
lhez
authored
clean up unused variables warnings (#23975)
anavp-nvidia
authored
server: real-time reasoning interruption via control endpoint (#23971)

ServeurpersoCom
and
allozaur
authored
hexagon: add gelu_quick (#24007)
tboinovski1
authored
hexagon: MUL_MAT, MUL_MAT_ID, FLASH_ATTN and GDN cleanup and optimizations for latest models (#23989)
max-krasnyansky
authored
llama : deprecate `llama_set_warmup` (#24009)

ggerganov
and
danbev
authored
convert : support Step3.7-Flash (#23845)

forforever73
and
CISC
authored
kv-cache : SWA checkpoints store only non-masked cells (#23981)
ggerganov
authored
ui: Add Thinking mode toggle with reasoning effort levels + improvements for Chat Form Add Action UI (#23434)
allozaur
authored
ui: simplify network error handling (#23431)
socram8888
authored
docs : update HOWTO-add-model.md (#23883)

Xarbirus
and
CISC
authored
ci : reduce self-hosted server workflow jobs (#24012)
ggerganov
authored
server: add SSE ping interval (#24013)
ngxson
authored
common : fix state save in common_prompt_batch_decode (#23468)

danbev
and
fairydreaming
authored
StepFun 3.5 MTP (#23274)

pwilkin
and
CISC
authored
model : support granite multilingual embeddings R2 (ibm-granite/granite-embedding-{97,311}m-multilingual-r2) (#22716)
hansolosan
authored
model: add Mellum architecture (#23966)

Xarbirus
and
CISC
authored
hexagon: profiler output fix and script updates (#24042)
max-krasnyansky
authored
opencl: use flat variants of q4_K and q6_K gemv for very large M (#24006)
lhez
authored

Commits on Jun 3, 2026

Commits on Jun 4, 2026

Commits on Jun 5, 2026

Commits on Jun 6, 2026