ggml-org/llama.cpp

Pull Requests Commits

reuse str_to_lower from server-models

ngxson committed 4 days ago

ddd9cdb1

server: (tools) add x-tool-cwd header

ngxson committed 5 days ago

9da93768

test: fix some CI errors (#26415)

ngxson committed 5 days ago

Verified 11924d4c

vulkan: extend topk_moe fusion to support sqrt(softplus) (#26124)

jeffbolznv committed 5 days ago

Verified a7a6d0d2

vendor : update BoringSSL to 0.20260730.0 (#26353)

cabelo committed 5 days ago

Verified 815a2a59

agents: clarify comment style and jinja knowledge (#26405)

ngxson committed 6 days ago

Verified 89482bd6

cli : persist reasoning_content in chat history (#26362)

ramicaza committed 6 days ago

Verified c629da56

mtmd: add minicpmv46 downsample (#25993)

tc-mb committed 6 days ago

Verified de699957

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

pwilkin committed 6 days ago

Verified ddd4ec14

vulkan: add POOL_1D op (#25431)

AnandPatil1 committed 7 days ago

Verified 876a4321

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

rillomas committed 7 days ago

Verified eb41d503

mtmd: add n_embd_head (#26342)

ngxson committed 7 days ago

Verified db7d8b24

Support rotated kv cache quant (#26180)

timkhronos committed 7 days ago

Verified a09d8abf

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

fairydreaming committed 7 days ago

Verified 82dbc4f0

vulkan: update vulkan sdk to 1.4.357.0 (#26303)

jeffbolznv committed 7 days ago

Verified 6f3c0a79

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

ruixiang63 committed 7 days ago

Verified 00054751

cuda: extract Q2_0 elements via __byte_perm (#25603)

dfriehs committed 7 days ago

Verified 15e755f3

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

johnkarlhill committed 7 days ago

Verified 9d9a6d29

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

arthw committed 7 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 7 days ago

Verified 69e62fc7

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

z-sachin committed 7 days ago

Verified 1e225995

sycl : support dev2dev memcpy by DEV2DEV_MEMCPY_FORWARD (#26234)

arthw committed 7 days ago

Verified 1c5b89ff

[SYCL] Support q2 mul_mat (#26231)

arthw committed 7 days ago

Verified a2be61dc

sycl: fuse RMS_NORM + MUL (#26015)

Titaniumtown committed 7 days ago

Verified 15537259

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

yomaytk committed 7 days ago

Verified 8f4646a6

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

ngxson committed 7 days ago

Verified 5f55650a

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

ngxson committed 7 days ago

Verified b4ca032a

vulkan: Support quantized concat (#25684)

jeffbolznv committed 8 days ago

Verified ea63b4d3

Test support for alternative conv layout (#25617)

pmaybank committed 8 days ago

Verified 958d9c0b

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

o7si committed 8 days ago

Verified 432d7ffe

Older