ngxson/llama.cpp

Pull Requests Commits

clip: fix nb calculation for qwen3-vl

ngxson committed 75 days ago

521516fb

ggml : add GGML_SCHED_NO_REALLOC option to disable reallocations in ggml_backend_sched (#17276)

slaren committed 76 days ago

Verified e072b205

[MUSA] enable fp16/fast_fp16/bf16_mma on PH1 (#17551)

yeahdongcn committed 76 days ago

Verified c6f7a423

ggml-cuda: add stricter checking for fusion (#17568)

am17an committed 76 days ago

Verified 2e7ef98f

server : add Anthropic Messages API support (#17570)

noname22 committed 76 days ago

Verified ddf9f943

model : Qwen3 Next (#16095)

pwilkin committed 76 days ago

Verified ff55414c

CUDA: no FP16 arithmetic for vector FA kernel (#17558)

JohannesGaessler committed 76 days ago

Verified 73955f7d

vulkan: Implement GGML_OP_TRI (#17503)

jeffbolznv committed 76 days ago

Verified 35cf8887

rpc : cache and reuse compute graphs (#15405)

rgerganov committed 76 days ago

Verified 15d2b46b

HIP: enable mul_mat_f for RDNA4 (#17437)

zhang-hui-yulo committed 76 days ago

Verified 6bca76ff

SOLVE_TRI CUDA kernel for small matrices (#17457)

pwilkin committed 76 days ago

Verified cd0e3a7a

refactor pad_reflect_1d to make the UT case pass (#17204)

NeoZhangJianyu committed 76 days ago

Verified efaaccdd

vulkan: Implement SOLVE_TRI (#17486)

jeffbolznv committed 77 days ago

Verified 4abef75f

arch : add description about LLM_TENSOR_INFOS (#17550)

ggerganov committed 77 days ago

Verified c3861149

models : fix LFM2 tensors (#17548)

ggerganov committed 77 days ago

Verified 6783b11f

cuda : fix UMA detection on discrete GPUs. (#17537)

matt23654 committed 77 days ago

Verified 909072ab

ggml-cpu: aarm64: q4_K repack gemm and gemv implementations (dotprod only) (#17494)

Alcpz committed 77 days ago

Verified cd8370b4

devops: Add build-essential to Ubuntu 26.04 image (#17531)

ericcurtin committed 77 days ago

Verified d21a76ac

gguf-py : skip endian-conversion of MXFP4 data (#17523)

AlekseiNikiforovIBM committed 77 days ago

Verified 4fcd87cf

vulkan : move contiguous checks to device_supports_op (#17490)

Acly committed 77 days ago

Verified b78db3bd

vulkan: use a fixed 1KB buffer for the add_rms_fusion opt (#17514)

jeffbolznv committed 77 days ago

Verified 142df17c

server: enable jinja by default, update docs (#17524)

ngxson committed 77 days ago

Verified e509411c

opencl: add sqr, sqrt, mean and ssm_conv (#17476)

lhez committed 77 days ago

Verified 7cba58bb

Fix chunks being too small with small matrix sizes (#17526)

Alcpz committed 77 days ago

Verified 5449367b

clip: (minicpmv) fix resampler kq_scale (#17516)

hNSBQZ committed 77 days ago

Verified 1d594c29

vulkan: allow graph_optimize for prompt processing workloads (#17475)

jeffbolznv committed 78 days ago

Verified eec1e33a

vulkan: Implement top-k (#17418)

jeffbolznv committed 78 days ago

Verified 879d6737

ggml-cpu : add RISC-V Zvfh impl for ggml_vec_mad_f16 (#17448)

xctan committed 78 days ago

Verified 6ab4e50d

cmake : use EXCLUDE_FROM_ALL to avoid patch-boringssl.cmake (#17520)

angt committed 78 days ago

Verified 2336cc47

ggml : fix ARM feature verification (#17519)

angt committed 78 days ago

Verified e6923caa

Older