Commits

History for

Commits on Jul 31, 2026

llama : load MTP tensors only if they are really used (#26296)

fairydreaming
and
sszymczy
authored
82dbc4f

Commits on Jul 27, 2026

mtmd: support MiMo-V2.5 audio input (RVQ-based model) (#26190)
ngxson
authored
1cbfd19

Commits on Jul 26, 2026

Keep Minimax's indexer tensors at F32 for speed and accuracy (#26144)
bartowski1182
authored
7657a6c

Commits on Jul 24, 2026

llama: various bug fixes (#26051)
ngxson
authored
298219f

Commits on Jul 23, 2026

args: refactor mlock/mmap/directio into load-mode (#20834)
taronaeo
authored
e6dd0e2

Commits on Jul 18, 2026

llama-quant : exclude i32 ffn_gate_tid2eid routing table from quantization (#25787)
devYRPauli
authored
4937ca8

Commits on Jul 16, 2026

quant : allow using manual tensor types with --pure (#25716)
ggerganov
authored
56d6e9d

Commits on Jul 7, 2026

Add Q2_0 quantization: type definition and CPU backend (#24448)
khosravipasha
authored
bec4772

Commits on Jun 25, 2026

quant : fix quantizing moe with mtp (#24986)
CISC
authored
b3ce5ce

Commits on Jun 20, 2026

llama : use LLM_KV for quantization_version & file_type (#24802)
angt
authored
84de01a

Commits on Jun 5, 2026

hparams : refactor `hparams.n_layer` (#24060)

ggerganov
and
CISC
authored
7acb4e8

Commits on May 4, 2026

model: move `load_hparams` and `load_tensors` to per-model definition (#22004)
ngxson
authored
994118a

Commits on May 1, 2026

llama-quant : fix `--tensor-type` when default `qtype` is overriden (#22572)
ddh0
authored
b97ebdc

Commits on Apr 25, 2026

llama-quant : default ftype param `Q5_1` --> `Q8_0` (#20828)
ddh0
authored
9d34231

Commits on Apr 6, 2026

ggml: add Q1_0 1-bit quantization support (CPU) (#21273)

khosravipasha
and
CISC
authored
2e1f0a8

Commits on Apr 2, 2026

tests : add unit test coverage for llama_tensor_get_type (#20112)

authored
7992aa7

Commits on Apr 1, 2026

llama : refactor llama_model_quantize_params to expose a pure C interface (#20346)

EAddario
and
ggerganov
authored
4951250

Commits on Mar 26, 2026

mtmd: fix "v.patch_embd" quant and unsupported im2col ops on Metal for deepseek-ocr (#21027)

sfallah
and
CISC
authored
1743d98

Commits on Mar 25, 2026

mtmd: Add DeepSeekOCR Support (#17400)

authored
a970515
llama: fix llama-model-saver (#20503)

JohannesGaessler
and
Siddhesh2377
authored
36dafba

Commits on Mar 10, 2026

Commits on Mar 9, 2026

llama-quant : left-align tensor names in output (#20117)
ddh0
authored
b518195

Commits on Mar 8, 2026

llama: end-to-end tests (#19802)
JohannesGaessler
authored
a976ff0

Commits on Feb 20, 2026

quantize : add --dry-run option (#19526)

ddh0
and
CISC
authored
492bc31

Commits on Feb 6, 2026

Kimi-Linear support (backend agnostic + MLA KV cache) (#18755)

ymcki
and
pwilkin
authored
3688c4f

Commits on Jan 28, 2026

llama : disable Direct IO by default (#19109)
ggerganov
authored
c5c64f7

Commits on Jan 22, 2026

quant : manual overrides of tensor types take precedence (#18952)
ggerganov
authored
0e4ebeb

Commits on Jan 8, 2026

llama : add `use_direct_io` flag for model loading (#18166)

authored
2038101

Commits on Dec 15, 2025

llama: automatically set parameters not set by the user in such a way that maximizes GPU utilization (#16653)
JohannesGaessler
authored
b1f3a6e

Commits on Dec 6, 2025

llama : remove quantization sanity check (#17788)
danbev
authored
444f00b

Commits on Dec 4, 2025

llama : fix sanity checks during quantization (#17721)
ggerganov
authored
a67ef0f

Commits on Nov 28, 2025

model : Qwen3 Next (#16095)

authored
ff55414

Commits on Oct 30, 2025

llama : use std::abs instead of abs (#16853)
kaetemi
authored
d739511

Commits on Oct 15, 2025

llama-quant: add support for mmproj (#16592)

ngxson
and
ggerganov
authored
3e3cb19