ggerganov/llama.cpp

Pull Requests Commits

rpc : add rdma_cm setup path for iWARP

rgerganov committed 6 days ago

a70de036

fix flush for apple rdma

am17an committed 6 days ago

d4052a4a

rpc: allow -sm tensor

am17an committed 6 days ago

89f02c4b

server : allow model downloads at model limit fix issue #26809 (#28530)

chenmiaoming committed 6 days ago

Verified 56381e40

ggml-cuda: hip add specific config table for AMD GCN (#27841)

thelittlefireman committed 7 days ago

Verified c8edceb0

server : add missing headers (#28795)

angt committed 7 days ago

Verified e192abb4

vendor : update cpp-httplib to 0.56.0 (#28787)

cabelo committed 7 days ago

Verified 718f7b41

syscl : Handle (fail gracefully) unsupported tq1_0 quants (#28681)

mctylr-gh committed 7 days ago

Verified 2a3005c2

rpc : fix linking when compiling with BUILD_SHARED_LIBS=OFF (#28492)

EAddario committed 7 days ago

Verified f3a33dff

server: frame the router child state command as a whole line (#28747)

ServeurpersoCom committed 7 days ago

Verified c069aa7f

opencl: fix several bugs where the backend aborts (#27630)

wanghqc committed 7 days ago

Verified 8a56aedd

opencl: add bin kernel `kernel_gemm_noshuffle_q4_k_f32_32b_trans_ila_a8_bin` (#28677)

shaofeiqi committed 7 days ago

Verified 07fc9771

webgpu: align tensor bindings to the type block size (#28382)

ServeurpersoCom committed 7 days ago

Verified 3f5e94d7

hexagon: support for multi-device model split (aka row-split) (#28589)

max-krasnyansky committed 7 days ago

Verified eafe15a5

ggml-webgpu: Update to a recent version of Dawn (#28683)

MendyBerger committed 7 days ago

Verified d3146f2b

server: refactor subproc handling (#28555)

ngxson committed 7 days ago

Verified 82d6bb28

cmake : skip PCH for llama-server PCH when using MSVC (#28763)

danbev committed 7 days ago

Verified 8ea29024

ci : run test-backend-ops as a dedicated ci/run.sh test (#28740)

ggerganov committed 7 days ago

Verified b78a39a2

tests: extend test-quantize-fns to test nrc=2 (i8mm) kernels (#16234)

Rohanjames1997 committed 7 days ago

Verified 982937a3

tests: tolerate a shared pool abort in test_completion_unified (#28759)

ServeurpersoCom committed 7 days ago

Verified 8172e657

ggml: skip 0-sized ids tensor when offloading selected experts (#28739)

am17an committed 7 days ago

Verified 43f3dda6

metal : skip the empty half of the mul_mm_id token tile (#28301)

masterFoad committed 7 days ago

Verified 5bda51bf

cmake : add PCH and unity build to improve build times (#28091)

danbev committed 7 days ago

Verified 3bcfeb70

common : fix typo in speculative.cpp comment [no ci] (#28750)

danbev committed 7 days ago

Verified 1dfe94e0

metal : single-source fusion table + fusion debug rework (#28164)

ggerganov committed 7 days ago

Verified a2878d30

metal : fix idle threads in the remaining iq mul_mv kernels for ne00 < 1024 (#28692)

masterFoad committed 8 days ago

Verified aac81023

model : fix MTP context kv cache allocation for deepseek2, glm4moe, … (#28630)

LoganChu committed 8 days ago

Verified 5cdd3d1d

server: fix speculation after an image (#28715)

jesdga95 committed 8 days ago

Verified b0dcb819

CUDA/HIP: Flash Attention tuning (gfx1201) (#28102)

pwilkin committed 8 days ago

Verified 16378d93

ci : key cache to sanitizer matrix (#28708)

CISC committed 8 days ago

Verified 451b89ba

Older