ngxson/llama.cpp

Pull Requests Commits

Merge branch 'master' into xsn/cmake_model_glob

ngxson committed 68 days ago

0f205e8b

model: using single llm_build per arch (#21970)

ngxson committed 68 days ago

Verified 4fbdabdc

opencl: add q5_K gemm and gemv kernels for Adreno (#21595)

shaofeiqi committed 68 days ago

Verified e45dbdec

server: tests: fetch random media marker via /apply-template (#21962) (#21980)

ServeurpersoCom committed 68 days ago

Verified 4adac43f

cmake: use glob to collect src/models sources

ngxson committed 68 days ago

a6495a50

model : refactor QKV into common build_qkv and create_tensor_qkv helpers (#21245)

JoursBleu committed 68 days ago

Verified 9db77a02

model : support NVFP4 tensors for Gemma4 (#21971)

CISC committed 68 days ago

Verified f772f6e4

codeowners: add team member comments (#21714)

0cc4m committed 68 days ago

Verified b572d1ec

Convert: Fix NemotronH Config Parsing (#21664)

anavp-nvidia committed 68 days ago

Verified 03b3d077

ggml: add graph_reused (#21764)

am17an committed 68 days ago

Verified 3f7c29d3

metal: Implement ROLL op (#21946)

kushagharahi committed 68 days ago

Verified ae2d3489

ggml-cpu: add 128-bit RVV implementation for Quantization Vector Dot (#20633)

rehan-10xengineer committed 68 days ago

Verified 1e796eb4

ggml : implemented simd_gemm kernel for riscv vector extension (#20627)

rehan-10xengineer committed 68 days ago

Verified 56375365

devops : added spirv-headers to nix (#21965)

yuannan committed 68 days ago

Verified 90fb96a7

ggml-webgpu: compute pass batching and removing profiling overhead (#21873)

reeselevine committed 68 days ago

Verified 82677a6e

ci : Use ggml-org/ccache-action on RISC-V as well (#21632)

luhenry committed 68 days ago

Verified 8612ed18

[SYCL] Fix Q8_0 reorder: garbage on 2nd prompt + crash on full VRAM (#21638)

PMZFX committed 68 days ago

Verified b1be68e8

server: use random media marker (#21962)

ngxson committed 68 days ago

Verified 408225bb

vulkan: optimize im2col (#21713)

0cc4m committed 69 days ago

Verified b3d75875

cuda: Q1_0 initial backend (#21629)

khosravipasha committed 69 days ago

Verified 7e72b38b

ggml-webgpu: Fix dequantization helpers to not pass in pointers (#21872)

reeselevine committed 69 days ago

Verified 20d3bc2c

CUDA: require explicit opt-in for P2P access (#21910)

JohannesGaessler committed 69 days ago

Verified a6206958

CUDA: manage NCCL communicators in context (#21891)

JohannesGaessler committed 69 days ago

Verified 014dca49

rpc : add native RDMA transport for RPC backend (RoCEv2) (#20590)

dvv101111 committed 69 days ago

Verified adb541a6

docs: more extensive RoPE documentation [no ci] (#21953)

ngxson committed 69 days ago

Verified 80d87708

ci: disable test-backend-ops on Vulkan llvmpipe run and resture default timeout (#21901)

0cc4m committed 69 days ago

Verified 8dc530b8

autoparser: support case of JSON_NATIVE with per-call markers (test case: Reka-Edge) (#21892)

pwilkin committed 69 days ago

Verified e1a9a6dc

read n_ctx back after making llama_context (#21939)

smashedpumpkin committed 69 days ago

Verified e39eba26

hexagon: optimization for HMX mat_mul (#21554)

njsyw1997 committed 69 days ago

Verified 5d14e5d1

ggml : remove ggml-ext.h (#21869)

ngxson committed 70 days ago

Verified fae3a280

Older