ggml-org/llama.cpp

Pull Requests Commits
rpc : fix regression when --device is used (#15981)
rgerganov committed 147 days ago
Verified 918b26f1
releases : update ROCM, add gfx1200, gfx1201, gfx1151 (#15972)
slaren committed 147 days ago
Verified 9ecb8843
doc : update documentation for --tensor-split (#15980)
rgerganov committed 147 days ago
Verified d1c6f11f
ggml-zdnn: rm user mapped buffers (#15965)
taronaeo committed 147 days ago
Verified 6380d6a3
vulkan: fix failing dequant shaders (#15862)
jeffbolznv committed 148 days ago
Verified aa0c461e
vulkan: initialize vulkan-hpp to allow using extension function pointers (#15705)
jeffbolznv committed 148 days ago
Verified b9c9c9f7
llama : allow using iGPUs with --device (#15951)
slaren committed 148 days ago
Verified 50f4281a
metal : refactor kernel loading (#15964)
ggerganov committed 148 days ago
Verified 55758b00
metal : allow ops to run concurrently (#15929)
ggerganov committed 148 days ago
Verified f161463a
metal : fix memory leaks (#15962)
ggerganov committed 148 days ago
Verified 84d7b2fc
ggml-zdnn: fix #15414, activate FP16 and BF16 acceleration and incorrect zTensor free (#15839)
taronaeo committed 149 days ago
Verified 40be5115
Add docker protocol support for llama-server model loading (#15790)
ericcurtin committed 149 days ago
Verified 4bf55492
context : remove redundant explicit casting to the same type (#15948)
haiyuewa committed 149 days ago
Verified f4e664f8
server : adjust prompt similarity thold + add logs (#15913)
ggerganov committed 149 days ago
Verified f088b6a8
Vulkan iGPU device selection overhaul and PCI ID API support (#15947)
0cc4m committed 149 days ago
Verified 304ac569
vulkan: Make device memory check more portable (#15939)
mbaudier committed 149 days ago
Verified 6c88ad8f
Revert "sycl: add usage of enqueue_functions extension (#14244)" (#15910)
NeoZhangJianyu committed 149 days ago
Verified 704d90c9
ggml-backend : add GGML_BACKEND_DEVICE_TYPE_IGPU device type (#15797)
slaren committed 150 days ago
Verified 360d6533
CUDA: larger SRAM reads for tile FA, AMD FP16 dot (#15927)
JohannesGaessler committed 150 days ago
Verified 0e6ff004
nitpick : correct MB to MiB (#15934)
ddh0 committed 150 days ago
Verified df082f56
ggml-cpu : add check for ARM MATMUL_INT8/i8mm support (#15922)
danbev committed 150 days ago
Verified 24a6734d
kleidiai: fix GGML_ASSERT(*cur_backend_id != -1) failed (#15614)
chaxu01 committed 150 days ago
Verified 2b3efea9
CANN: Disable acl_graph for prefill stage (#15933)
hipudding committed 150 days ago
Verified c0389dba
CUDA: Add `fastdiv` to `k_bin_bcast*`, giving 1-3% E2E performance (#15872)
ORippler committed 151 days ago
Verified 00681dfc
llama : support T5 models with unequal number of encoder-decoder layers (#15909)
DamonFool committed 151 days ago
Verified 4f658855
graph : support non-contiguous Q in build_attn_mha (#15908)
CISC committed 151 days ago
Verified 6ab397e1
ggml-cpu : fix padding in ggml_timestep_embedding (#15917)
danbev committed 151 days ago
Verified 9de447d9
metal : make the backend async (#15906)
ggerganov committed 151 days ago
Verified 0f0a3c28
ci : add caching for ROCm installation in release workflow (#15924)
danbev committed 151 days ago
Verified 33daece8
tests : filter out no-ops from coverage report (#15900)
danbev committed 151 days ago
Verified e7b6d83b
Terms Of Use Privacy Policy Refund Policy Impressum