ngxson/llama.cpp

Pull Requests Commits

memory: add is_iswa for memory_hybrid

ngxson committed 24 days ago

d3e33220

gguf-py : add requests to dependencies (#18629)

ckastner committed 24 days ago

Verified bd2a93d4

ggml : fix avx512bf16 build (#18623)

angt committed 24 days ago

Verified e75ee110

CANN: Make `valid_values` variable `static const` (#18627)

rauletorresc committed 24 days ago

Verified da9b8d33

ggml webgpu: add CEIL operation support (#18605)

tnguyen21 committed 24 days ago

Verified e443fbcf

model : add LFM2-ColBert-350M (#18607)

tdakhran committed 24 days ago

Verified 73d284a2

CUDA: fix FA FP16 accumulator overflow for Granite (#18614)

JohannesGaessler committed 24 days ago

Verified df17a4c9

add YoutuVLForConditionalGeneration architectures (#18620)

f291400 committed 24 days ago

Verified 1871f0ba

ggml-cuda: check for srcs outside the cgraph (#18583)

am17an committed 24 days ago

Verified f47edb8c

server : fix router child env in containerized environments (#18562)

v-sayapin committed 25 days ago

Verified da143b99

vulkan: fix topk_moe_sigmoid_norm_bias failures in GLM-4.6 (#18582)

jeffbolznv committed 25 days ago

Verified f1768d8f

models : fix backend assignment for Granite/Nemotron graphs (#18599)

ggerganov committed 25 days ago

Verified 2da64a2f

vulkan: handle quantize_q8_1 overflowing the max workgroup count (#18515)

jeffbolznv committed 25 days ago

Verified b37124d2

llama : refactor rope_freq_base/scale_swa conversion and init (#18553)

CISC committed 25 days ago

Verified eadc4184

CANN: add operator fusion support for ADD + RMS_NORM (#17512)

noemotiovon committed 25 days ago

Verified 67e3f6f6

doc: clarify that steps also apply to linux for opencl (#18002)

alosslessdev committed 25 days ago

Verified 92ac1e01

ci : init git lfs in every build for RISC-V (#18590)

alitariq4589 committed 25 days ago

Verified 8e3a7611

sampling : add support for backend sampling (#17004)

danbev committed 25 days ago

Verified d3dce4e0

model : mtmd : make input norm optional in LFM2-VL (#18594)

tdakhran committed 25 days ago

Verified 4974bf53

CUDA: disable cuda graph when using n-cpu-moe (#18593)

am17an committed 25 days ago

Verified 908a9e5a

ggml-cuda: remove unused params in ggml_cuda_graph (#18579)

am17an committed 25 days ago

Verified 5126c41c

common/grammar : replace problematic backtracking regex `[\s\S]*` (#18342)

aldehir committed 26 days ago

Verified cef1d23c

graph : fix graph reuse logic when `n_pos_per_embd > 1` (#18566)

ggerganov committed 26 days ago

Verified c69c7ebc

ggml-cuda: fixes for concurrent streams (#18496)

am17an committed 26 days ago

Verified e57f5233

context : fix reserve token padding to n_seqs (#18536)

ggerganov committed 27 days ago

Verified a554a1ec

CUDA: only allocate FA tmp buffer if needed (#18564)

JohannesGaessler committed 27 days ago

Verified 0f2e42ca

(Bugfix, ggml-cuda) Pool alloc count fix + small size computation type adjustment (#18559)

pl752 committed 27 days ago

Verified 9dba9f53

ggml-hexagon: optimize activation function (#18393)

joeldushouyu committed 27 days ago

Verified bcfc8c3c

vulkan: Optimize GGML_OP_CUMSUM (#18417)

jeffbolznv committed 27 days ago

Verified 18ddaea2

vulkan: Implement mmvq for iq1_s/iq1_m (#18450)

jeffbolznv committed 27 days ago

Verified 706e3f93

Older