5 people 153d324bcf llama: add default load-mode auto, which avoids mmap on iGPUs (#26081) ...

* llama: add new default load-mode auto which picks mmap unless a non-Metal iGPU is used

* Update ggml/src/ggml-hexagon/ggml-hexagon.cpp

Co-authored-by: Max Krasnyansky <maxk@qti.qualcomm.com>

* set mmap_support to false on OpenCL backend

* fix order of load modes

* use -1 for auto

* resolve load mode auto earlier to correctly pick gpu host or cpu memory

* add load mode auto to llama-bench

* bump virtgpu api version, regenerate docs

---------

Co-authored-by: Piotr Wilkin (ilintar) <piotr.wilkin@syndatis.com>
Co-authored-by: Max Krasnyansky <maxk@qti.qualcomm.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

2026-08-11 09:20:46 +03:00

..

batched-bench

cmake : add install() for impl libraries + fix apple builds (#23511)

2026-05-22 11:46:26 +03:00

cli

llama: add default load-mode auto, which avoids mmap on iGPUs (#26081)

2026-08-11 09:20:46 +03:00

completion

llama: add default load-mode auto, which avoids mmap on iGPUs (#26081)

2026-08-11 09:20:46 +03:00

cvector-generator

libs : rename libcommon -> libllama-common (#21936)

2026-04-17 11:11:46 +03:00

export-lora

docs: fix export-lora --lora-scaled syntax [no release] (#24703)

2026-06-18 16:46:17 +02:00

fit-params

cmake : add install() for impl libraries + fix apple builds (#23511)

2026-05-22 11:46:26 +03:00

gguf-split

gguf-split: Add option to delete split parts during merge (#26538)

2026-08-04 21:27:47 +03:00

imatrix

Move duplicated imatrix code into single common imatrix-loader.cpp (#22445)

2026-06-04 17:45:40 +02:00

llama-bench

llama: add default load-mode auto, which avoids mmap on iGPUs (#26081)

2026-08-11 09:20:46 +03:00

mtmd

model: Muse Glimmer Support (#26841)

2026-08-10 13:07:27 +02:00

parser

chat : enable tool call in thinking for DS4 (#26269)

2026-08-01 00:13:07 -05:00

perplexity

perplexity : fix format specifier in LOG_ERR (#23788)

2026-05-28 10:34:58 +03:00

quantize

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

2026-07-07 12:05:47 -07:00

results

libs : rename libcommon -> libllama-common (#21936)

2026-04-17 11:11:46 +03:00

rpc

binaries : Improve rpc-server and export-graph-ops names. (#25045)

2026-06-27 10:31:29 +03:00

server

llama: add default load-mode auto, which avoids mmap on iGPUs (#26081)

2026-08-11 09:20:46 +03:00

tokenize

tokenize : drop --stdin mutual-exclusion check (#25672)

2026-07-15 18:41:51 +02:00

tts

tts: account for the vocoder pass in the timings line (#26733)

2026-08-07 22:35:52 +02:00

ui

ui: UI/chat form follow ups (#26743)

2026-08-10 13:32:51 +02:00

CMakeLists.txt

cmake: skip cvector-generator and export-lora when CPU backend is disabled (#24053)

2026-06-04 13:13:19 +03:00