ggml-org/llama.cpp

Pull Requests Commits

improve Security review a bit

ngxson committed 23 days ago

37e22b75

agents: clarify comment style and jinja knowledge

ngxson committed 23 days ago

b2e89532

mtmd: add minicpmv46 downsample (#25993)

tc-mb committed 23 days ago

Verified de699957

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

pwilkin committed 24 days ago

Verified ddd4ec14

vulkan: add POOL_1D op (#25431)

AnandPatil1 committed 24 days ago

Verified 876a4321

vulkan: Introduce driver version check for Windows Intel GPU to mitigate crashing (#25192)

rillomas committed 24 days ago

Verified eb41d503

mtmd: add n_embd_head (#26342)

ngxson committed 24 days ago

Verified db7d8b24

Support rotated kv cache quant (#26180)

timkhronos committed 24 days ago

Verified a09d8abf

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

fairydreaming committed 24 days ago

Verified 82dbc4f0

vulkan: update vulkan sdk to 1.4.357.0 (#26303)

jeffbolznv committed 24 days ago

Verified 6f3c0a79

server: correct accepted tokens when need draft token replay (#26320)

ruixiang63 committed 25 days ago

Verified 00054751

cuda: extract Q2_0 elements via __byte_perm (#25603)

dfriehs committed 25 days ago

Verified 15e755f3

SYCL: add oneMKL GEMM flash attention for XMX-accelerated prompt proc… (#25025)

johnkarlhill committed 25 days ago

Verified 9d9a6d29

[SYCL] support the missed types in cpy (#26005)

arthw committed 25 days ago

Verified d5d3e05b

llama : enforce the same K and V cache types for DeepSeek V4; enable FA if V cache is quantized (#25871)

fairydreaming committed 25 days ago

Verified 69e62fc7

ggml-zendnn : group matmul direct API for mul_mat_id (#25918)

z-sachin committed 25 days ago

Verified 1e225995

sycl : support dev2dev memcpy by DEV2DEV_MEMCPY_FORWARD (#26234)

arthw committed 25 days ago

Verified 1c5b89ff

[SYCL] Support q2 mul_mat (#26231)

arthw committed 25 days ago

Verified a2be61dc

sycl: fuse RMS_NORM + MUL (#26015)

Titaniumtown committed 25 days ago

Verified 15537259

ggml-webgpu: improve flash_attn_vec for quantized KV at long contexts (#25956)

yomaytk committed 25 days ago

Verified 8f4646a6

mtmd: add lanczos resize method [no release] (#26341)

ngxson committed 25 days ago

Verified 5f55650a

server: support inp embd to generate next token (#26313)

ngxson committed 25 days ago

Verified b4ca032a

vulkan: Support quantized concat (#25684)

jeffbolznv committed 25 days ago

Verified ea63b4d3

Test support for alternative conv layout (#25617)

pmaybank committed 25 days ago

Verified 958d9c0b

llama-context : sync pending async copies before clearing embd_seq (#25676)

o7si committed 25 days ago

Verified 432d7ffe

tests : avoid building get-model.cpp many times (#26317)

ggerganov committed 25 days ago

Verified 47f686f5

ggml-cuda: Allow transpose-free gemmv computation (#26171)

roberteg16 committed 25 days ago

Verified e1a1abb7

readme : refresh (#26280)

ggerganov committed 25 days ago

Verified 6b36c230

ggerganov committed 25 days ago

9ebfc3a8

ggml : bump version to 0.18.0 (ggml/1576)

ggerganov committed 25 days ago

6a4c3357

Older