s-goto-11
|
078ce23ea7
ggml: SVE support for exponential functions (#15145)
|
пре 4 месеци |
Prashant Vithule
|
a0c2b207c5
ggml: aarch64: Implement SVE F16 kernels for vector functions (#15115)
|
пре 4 месеци |
Jie Fu (傅杰)
|
4b20d8b7e3
convert : remove redundant code (#15708)
|
пре 4 месеци |
Ruben Ortlam
|
02c1813517
Vulkan: Add Integer Dot Product mul_mat_vec shader for legacy quants (#14903)
|
пре 4 месеци |
Daniel Bevenius
|
77dee9de97
ggml : WebGPU add TRANSPOSE and RESHAPE to supported ops (#15695)
|
пре 4 месеци |
Jie Fu (傅杰)
|
4795c91c32
docs : add Hunyuan to models section (#15707)
|
пре 4 месеци |
Akarshan Biswas
|
b66df9d9c9
CUDA: fix build error from ambiguous __half conversions in conv2d (#15690)
|
пре 4 месеци |
hipudding
|
b9382c3877
CANN: Optimize MUL_MAT_ID (#15658)
|
пре 4 месеци |
hipudding
|
3dc7397a27
CANN: fix RoPE cache issue on multi-device (#15629)
|
пре 4 месеци |
Georgi Gerganov
|
e92d53b29e
sampling : optimize samplers by reusing bucket sort (#15665)
|
пре 4 месеци |
Georgi Gerganov
|
0d161f021a
server : enable /slots by default and make it secure (#15630)
|
пре 4 месеци |
Georgi Gerganov
|
4efd5a8316
metal : fix checks for available FA kernels (#15700)
|
пре 4 месеци |
Diego Devesa
|
274966226f
llama : fix fattn reserve call n_seqs parameter (#15699)
|
пре 4 месеци |
Diego Devesa
|
9777032dcc
llama : separate compute buffer reserve from fattn check (#15696)
|
пре 4 месеци |
Sigbjørn Skjæret
|
7d3c9f2b21
ci : explicitly set fa off or on (#15692)
|
пре 4 месеци |
Jeff Bolz
|
bbbf5ecccb
vulkan: handle large sizes for get_rows (#15686)
|
пре 4 месеци |
Jeff Bolz
|
c37052ab4d
vulkan: mul_mat_id coopmat2 optimizations (#15546)
|
пре 4 месеци |
Daniel Bevenius
|
5c16b9c87d
vulkan : remove unused portability_enumeration_ext variable (#15679)
|
пре 4 месеци |
Jeff Bolz
|
b97c9edc59
vulkan: Allow fallback to sysmem memory when vidmem is full (#15649)
|
пре 4 месеци |
Jeff Bolz
|
94e82c7ead
vulkan: clamp matmul and FA results to the max finite value (#15652)
|
пре 4 месеци |
Charles Xu
|
4d74393bcc
ggml: update kleidiai to v1.13.0 (#15663)
|
пре 4 месеци |
Diego Devesa
|
dd892555b0
Update build.md to remove MSVC arm64 notes (#15684)
|
пре 4 месеци |
Johannes Gäßler
|
e81b8e4b7f
llama: use FA + max. GPU layers by default (#15434)
|
пре 4 месеци |
Johannes Gäßler
|
38ad381f9f
CUDA: use FP32 arithmetic for conv2d (#15683)
|
пре 4 месеци |
Jeff Bolz
|
696fccf354
vulkan: Skip syncing for prealloc_y when it is reused (#15544)
|
пре 4 месеци |
Chenguang Li
|
ef476916bb
CANN: FIx compiler warnings (#15661)
|
пре 4 месеци |
Sergey Alirzaev
|
d82f6aa34a
server : removed obsolete doc (#15670)
|
пре 4 месеци |
Johannes Gäßler
|
3d16b29c3b
scripts: strip "AMD Instinct" from GPU name (#15668)
|
пре 4 месеци |
ExtReMLapin
|
792b44f2ed
server : add documentation for `parallel_tool_calls` param (#15647)
|
пре 4 месеци |
Aman Gupta
|
81017865ee
CUDA: fix bug in rms_norm fusion (#15660)
|
пре 4 месеци |