ggml-org/llama.cpp

Pull Requests Commits

CANN: Fix type float_t to float (#15736)

noemotiovon committed 187 days ago

Verified 8a2234ea

fix: resolve unsigned int initialization warning for n_dims/size in gguf.cpp (#15754)

skrandy committed 187 days ago

Verified 3de00820

chore: Update `.clang-format` to use `BinPackArguments=true` (#15744)

ORippler committed 187 days ago

Verified 69db8a52

llama: -fa 1/0/-1 aliases for -fa on/off/auto (#15746)

JohannesGaessler committed 187 days ago

Verified c466abe1

vulkan: fix shaders gen when no integer dot is available (#15740)

0cc4m committed 187 days ago

Verified 0a2a3841

CANN: Resolve soft_max precision issue (#15730)

hipudding committed 187 days ago

Verified 9961d244

vulkan: Fix macro parameter order for f32 matmul shaders (#15716)

jeffbolznv committed 188 days ago

Verified 25f1045f

opencl: add attn sinks support for FA kernels (#15706)

rmatif committed 188 days ago

Verified 97669e40

CANN: Support eager execution mode under ACL graph compilation (#15712)

noemotiovon committed 188 days ago

Verified 2f853687

CANN: Support ext_factor in rope (#15710)

hipudding committed 188 days ago

Verified ef2af57d

ggml-backend: raise GGML_MAX_SPLIT_INPUTS (#15722)

JohannesGaessler committed 188 days ago

Verified 5d804a49

vulkan: use memory budget extension to read memory usage (#15545)

giladgd committed 188 days ago

Verified d4d8dbe3

vulkan: add missing clamps in new mul_mat_id paths (#15702)

jeffbolznv committed 188 days ago

Verified 35a42eda

vulkan: disable large mmv subgroups on older Nvidia GPUs (#15717)

0cc4m committed 188 days ago

Verified fec7911f

ggml: SVE support for exponential functions (#15145)

s-goto-11 committed 188 days ago

Verified 078ce23e

ggml: aarch64: Implement SVE F16 kernels for vector functions (#15115)

Vithulep committed 188 days ago

Verified a0c2b207

convert : remove redundant code (#15708)

DamonFool committed 188 days ago

Verified 4b20d8b7

Vulkan: Add Integer Dot Product mul_mat_vec shader for legacy quants (#14903)

0cc4m committed 188 days ago

Verified 02c18135

ggml : WebGPU add TRANSPOSE and RESHAPE to supported ops (#15695)

danbev committed 188 days ago

Verified 77dee9de

docs : add Hunyuan to models section (#15707)

DamonFool committed 188 days ago

Verified 4795c91c

CUDA: fix build error from ambiguous __half conversions in conv2d (#15690)

qnixsynapse committed 189 days ago

Verified b66df9d9

CANN: Optimize MUL_MAT_ID (#15658)

hipudding committed 189 days ago

Verified b9382c38

CANN: fix RoPE cache issue on multi-device (#15629)

hipudding committed 189 days ago

Verified 3dc7397a

sampling : optimize samplers by reusing bucket sort (#15665)

ggerganov committed 189 days ago

Verified e92d53b2

server : enable /slots by default and make it secure (#15630)

ggerganov committed 189 days ago

Verified 0d161f02

metal : fix checks for available FA kernels (#15700)

ggerganov committed 189 days ago

Verified 4efd5a83

llama : fix fattn reserve call n_seqs parameter (#15699)

slaren committed 189 days ago

Verified 27496622

llama : separate compute buffer reserve from fattn check (#15696)

slaren committed 189 days ago

Verified 9777032d

ci : explicitly set fa off or on (#15692)

CISC committed 189 days ago

Verified 7d3c9f2b

vulkan: handle large sizes for get_rows (#15686)

jeffbolznv committed 189 days ago

Verified bbbf5ecc

Newer Older