ngxson/llama.cpp

Pull Requests Commits

CUDA: support for weight clamp in top-k norm (#16702)

am17an committed 72 days ago

Verified 75d33b93

ggml-alloc : make gallocr prefer chunks that allow memory reuse (#16788)

Acly committed 72 days ago

Verified 3470a5c8

cuda : use fast copy when src and dst are of different type and contiguous (#16789)

CISC committed 72 days ago

Verified bd562fe4

ggml: fix cuda kernel launch configuration for k_compute_batched_ptrs to support large batch (#16744)

leejet committed 72 days ago

Verified bbac6a26

convert : enable expert group selection for all models with it (#16691)

CISC committed 72 days ago

Verified 73a48c97

graph : add clamping to ffn_moe_weights_sum to avoid div-by-zero (#16655)

CISC committed 72 days ago

Verified f696428c

model : set res->t_embd in SmallThinker models (#16782)

CISC committed 72 days ago

Verified 7cce4f81

docs : add Jamba to Text-only models list (#16778)

amirai21 committed 72 days ago

Verified 8d886282

CUDA: General GEMV fusion (#16715)

am17an committed 72 days ago

Verified f77c13b9

vulkan: deduplicate Microsoft Direct3D12 devices (#16689)

giladgd committed 73 days ago

Verified 3cfa9c3f

convert : handle mmproj filename/path properly (#16760)

Galunid committed 73 days ago

Verified 5d195f17

model : set res->t_embd in PLaMo2 models (#16766)

mitmul committed 73 days ago

Verified 226f295f

vulkan: delete dead code (#16732)

giuseppe committed 74 days ago

Verified f90b4a8e

vulkan: Optimize SSM_SCAN (#16645)

jeffbolznv committed 74 days ago

Verified 8423d019

convert : avoid dequantizing mxfp4 for GPT-OSS (#16756)

compilade committed 74 days ago

Verified 5cca2542

ggml: fix CUDA grid launch condition for large block_nums.y in binbcast (#16742)

leejet committed 74 days ago

Verified 55945d2e

CUDA: use CUB for arbitary size argsort (#16754)

am17an committed 74 days ago

Verified 0bcb40b4

webui: support q URL parameter (#16728)

odrling committed 74 days ago

Verified 69e9ff01

model-conversion : add trust_remote_code for orig model run [no ci] (#16751)

danbev committed 74 days ago

Verified 5a91109a

convert : handle pre-quantized models (#14810)

compilade committed 75 days ago

Verified f8f071fa

server: add memory breakdown print (#16740)

JohannesGaessler committed 75 days ago

Verified 0bf47a1d

convert : Make mistral-common dependency optional (#16738)

juliendenize committed 75 days ago

Verified dd62dcfa

mtmd-cli : allow using --jinja (#16718)

ngxson committed 75 days ago

Verified d0660f23

Manually link -lbsd to resolve flock symbol on AIX (#16610)

mehendarkarprajwal committed 75 days ago

Verified fe6a9882

ggml-cuda: use passed ops instead of hardcoded ops (#16712)

am17an committed 75 days ago

Verified 061f0eff

server : send partial stop string when <EOG> is reached (#15007)

matteoserva committed 75 days ago

Verified 8cf6b42d

sycl: use async memory allocation to fix crashes during graph recording (#16644)

mmichel11 committed 76 days ago

Verified 9de9672a

Add experimental ggml-hexagon backend for the Hexagon NPU (#16547)

max-krasnyansky committed 76 days ago

Verified 63d2fc46

Revert "ggml : Leverage the existing GGML_F32_VEC helpers to vectorize ggml_v…" (#16723)

slaren committed 76 days ago

Verified a2e0088d

webui: introduce OpenAI-compatible model selector in JSON payload (#16562)

ServeurpersoCom committed 76 days ago

Verified 9b9201f6

Newer Older