Georgi Gerganov
|
3120413ccd
vulkan : remove unused vars (#0)
|
6 сар өмнө |
Acly
|
74bb294591
vulkan : implement bilinear interpolation (ggml/1291)
|
6 сар өмнө |
Acly
|
3e303b1107
vulkan : implement ggml_roll (ggml/1290)
|
6 сар өмнө |
Jeff Bolz
|
b3ad3a0191
vulkan: support SET_ROWS (#14587)
|
6 сар өмнө |
Jeff Bolz
|
98197e5c98
vulkan: optimizations for deepseek prompt processing (#14555)
|
6 сар өмнө |
Xuan-Son Nguyen
|
98bab638fb
ggml : add ggml_scale_bias (#14417)
|
6 сар өмнө |
Jeff Bolz
|
6efcd65945
vulkan: optimize flash attention split_k_reduce (#14554)
|
6 сар өмнө |
Jeff Bolz
|
e592be1575
vulkan: fix rms_norm+mul fusion (#14545)
|
6 сар өмнө |
Jeff Bolz
|
a0374a67e2
vulkan: Handle updated FA dim2/3 definition (#14518)
|
6 сар өмнө |
Sigbjørn Skjæret
|
28657a8229
ggml : implement GEGLU_ERF and GEGLU_QUICK ops (#14445)
|
6 сар өмнө |
Jeff Bolz
|
2b72bedec1
vulkan: support mixed/deepseekR1 FA head sizes (#14509)
|
6 сар өмнө |
Georgi Gerganov
|
a70c8a0c4b
kv-cache : use ggml_set_rows (#14285)
|
6 сар өмнө |
Georgi Gerganov
|
9067487c44
ggml : fix FA mask dim 2 and 3 (#14505)
|
6 сар өмнө |
Jeff Bolz
|
8875523eb3
vulkan: support softmax/FA batch and broadcast (#14449)
|
7 сар өмнө |
Georgi Gerganov
|
ec68e84c32
ggml : support bcast ggml_soft_max_ext, ggml_flash_attn_ext (#14435)
|
7 сар өмнө |
Jeff Bolz
|
6a746cf9c4
vulkan: Split large mul_mat_id to fit in shared memory (#14451)
|
7 сар өмнө |
Sigbjørn Skjæret
|
eff5e45443
add GELU_ERF (#14455)
|
7 сар өмнө |
Sigbjørn Skjæret
|
a0535ffa0d
ggml : implement REGLU/GEGLU/SWIGLU ops (#14158)
|
7 сар өмнө |
Jeff Bolz
|
bd9c981d72
vulkan: Add fusion support for RMS_NORM+MUL (#14366)
|
7 сар өмнө |
Jeff Bolz
|
63a7bb3c7e
vulkan: handle noncontig in the final case of ggml_vk_get_cpy_pipeline (#14378)
|
7 сар өмнө |
Jeff Bolz
|
00d5282c7f
vulkan: lock accesses of pinned_memory vector (#14333)
|
7 сар өмнө |
Markus Tavenrath
|
bb16041cae
Add support for VK_EXT_debug_utils to add labels to Vulkan objects. (#13792)
|
7 сар өмнө |
0cc4m
|
10bb545c5b
Vulkan: Set device max size for host memory to avoid OOM warning and fallback to CPU buffer (#14249)
|
7 сар өмнө |
Jeff Bolz
|
c89c2d1ab9
vulkan: mutex around vkQueueSubmit (#14127)
|
7 сар өмнө |
Jeff Bolz
|
bd248d4dc7
vulkan: Better thread-safety for command pools/buffers (#14116)
|
7 сар өмнө |
Jeff Bolz
|
1f7d50b293
vulkan: Track descriptor pools/sets per-context (#14109)
|
7 сар өмнө |
0cc4m
|
97340b4c99
Vulkan: Don't default to CPU device (like llvmpipe), even if no other device is available, to allow fallback to CPU backend (#14099)
|
7 сар өмнө |
Masato Nakasaka
|
669c13e0f6
vulkan: Enable VK_KHR_cooperative_matrix extension for Intel Xe2 GPUs (#14001)
|
7 сар өмнө |
Jeff Bolz
|
5a8ae3053c
vulkan: automatically deduce size of push constants (#13936)
|
7 сар өмнө |
Ervin Áron Tasnádi
|
0d3984424f
ggml-vulkan: adds support for op CONV_TRANSPOSE_1D (#13813)
|
7 сар өмнө |