| .. |
|
acc.cu
|
ae1f211ce2
cuda : refactor into multiple files (#6269)
|
hai 1 ano |
|
acc.cuh
|
ae1f211ce2
cuda : refactor into multiple files (#6269)
|
hai 1 ano |
|
arange.cu
|
ae1f211ce2
cuda : refactor into multiple files (#6269)
|
hai 1 ano |
|
arange.cuh
|
ae1f211ce2
cuda : refactor into multiple files (#6269)
|
hai 1 ano |
|
argsort.cu
|
08a0c02060
ggml : mul_mat_id use the same tensor for all the experts (#6387)
|
hai 1 ano |
|
argsort.cuh
|
ae1f211ce2
cuda : refactor into multiple files (#6269)
|
hai 1 ano |
|
binbcast.cu
|
0d56246f4b
ggml : group all experts in a single ggml_mul_mat_id (#6505)
|
hai 1 ano |
|
binbcast.cuh
|
ae1f211ce2
cuda : refactor into multiple files (#6269)
|
hai 1 ano |
|
clamp.cu
|
bc4bba364f
Introduction of CUDA Graphs to LLama.cpp (#6766)
|
hai 1 ano |
|
clamp.cuh
|
ae1f211ce2
cuda : refactor into multiple files (#6269)
|
hai 1 ano |
|
common.cuh
|
133d99c599
CUDA: deduplicate FlashAttention code (#7352)
|
hai 1 ano |
|
concat.cu
|
ae1f211ce2
cuda : refactor into multiple files (#6269)
|
hai 1 ano |
|
concat.cuh
|
ae1f211ce2
cuda : refactor into multiple files (#6269)
|
hai 1 ano |
|
convert.cu
|
e84b71c2c6
ggml : drop support for QK_K=64 (#7473)
|
hai 1 ano |
|
convert.cuh
|
5dc9dd7152
llama : add Command R Plus support (#6491)
|
hai 1 ano |
|
cpy.cu
|
bc4bba364f
Introduction of CUDA Graphs to LLama.cpp (#6766)
|
hai 1 ano |
|
cpy.cuh
|
bc4bba364f
Introduction of CUDA Graphs to LLama.cpp (#6766)
|
hai 1 ano |
|
dequantize.cuh
|
5dc9dd7152
llama : add Command R Plus support (#6491)
|
hai 1 ano |
|
diagmask.cu
|
ae1f211ce2
cuda : refactor into multiple files (#6269)
|
hai 1 ano |
|
diagmask.cuh
|
ae1f211ce2
cuda : refactor into multiple files (#6269)
|
hai 1 ano |
|
dmmv.cu
|
e84b71c2c6
ggml : drop support for QK_K=64 (#7473)
|
hai 1 ano |
|
dmmv.cuh
|
d48ccf3ad4
sync : ggml (#6351)
|
hai 1 ano |
|
fattn-common.cuh
|
133d99c599
CUDA: deduplicate FlashAttention code (#7352)
|
hai 1 ano |
|
fattn-tile-f16.cu
|
cd93a28cb1
CUDA: fix FA out-of-bounds reads (#7479)
|
hai 1 ano |
|
fattn-tile-f16.cuh
|
0fc1e820a9
CUDA: faster large batch FA without tensor cores (#7314)
|
hai 1 ano |
|
fattn-tile-f32.cu
|
cd93a28cb1
CUDA: fix FA out-of-bounds reads (#7479)
|
hai 1 ano |
|
fattn-tile-f32.cuh
|
0fc1e820a9
CUDA: faster large batch FA without tensor cores (#7314)
|
hai 1 ano |
|
fattn-vec-f16.cu
|
cd93a28cb1
CUDA: fix FA out-of-bounds reads (#7479)
|
hai 1 ano |
|
fattn-vec-f16.cuh
|
dc685be466
CUDA: add FP32 FlashAttention vector kernel (#7188)
|
hai 1 ano |
|
fattn-vec-f32.cu
|
cd93a28cb1
CUDA: fix FA out-of-bounds reads (#7479)
|
hai 1 ano |
|
fattn-vec-f32.cuh
|
dc685be466
CUDA: add FP32 FlashAttention vector kernel (#7188)
|
hai 1 ano |
|
fattn.cu
|
133d99c599
CUDA: deduplicate FlashAttention code (#7352)
|
hai 1 ano |
|
fattn.cuh
|
9c67c2773d
ggml : add Flash Attention (#5021)
|
hai 1 ano |
|
getrows.cu
|
ae1f211ce2
cuda : refactor into multiple files (#6269)
|
hai 1 ano |
|
getrows.cuh
|
ae1f211ce2
cuda : refactor into multiple files (#6269)
|
hai 1 ano |
|
im2col.cu
|
ae1f211ce2
cuda : refactor into multiple files (#6269)
|
hai 1 ano |
|
im2col.cuh
|
ae1f211ce2
cuda : refactor into multiple files (#6269)
|
hai 1 ano |
|
mmq.cu
|
e84b71c2c6
ggml : drop support for QK_K=64 (#7473)
|
hai 1 ano |
|
mmq.cuh
|
ae1f211ce2
cuda : refactor into multiple files (#6269)
|
hai 1 ano |
|
mmvq.cu
|
bc4bba364f
Introduction of CUDA Graphs to LLama.cpp (#6766)
|
hai 1 ano |
|
mmvq.cuh
|
ae1f211ce2
cuda : refactor into multiple files (#6269)
|
hai 1 ano |
|
norm.cu
|
ae1f211ce2
cuda : refactor into multiple files (#6269)
|
hai 1 ano |
|
norm.cuh
|
ae1f211ce2
cuda : refactor into multiple files (#6269)
|
hai 1 ano |
|
pad.cu
|
ae1f211ce2
cuda : refactor into multiple files (#6269)
|
hai 1 ano |
|
pad.cuh
|
ae1f211ce2
cuda : refactor into multiple files (#6269)
|
hai 1 ano |
|
pool2d.cu
|
ae1f211ce2
cuda : refactor into multiple files (#6269)
|
hai 1 ano |
|
pool2d.cuh
|
ae1f211ce2
cuda : refactor into multiple files (#6269)
|
hai 1 ano |
|
quantize.cu
|
5dc9dd7152
llama : add Command R Plus support (#6491)
|
hai 1 ano |
|
quantize.cuh
|
5dc9dd7152
llama : add Command R Plus support (#6491)
|
hai 1 ano |
|
rope.cu
|
3e5faa8503
cuda : fix rope + add tests (#7452)
|
hai 1 ano |
|
rope.cuh
|
ae1f211ce2
cuda : refactor into multiple files (#6269)
|
hai 1 ano |
|
scale.cu
|
bc4bba364f
Introduction of CUDA Graphs to LLama.cpp (#6766)
|
hai 1 ano |
|
scale.cuh
|
ae1f211ce2
cuda : refactor into multiple files (#6269)
|
hai 1 ano |
|
softmax.cu
|
133d99c599
CUDA: deduplicate FlashAttention code (#7352)
|
hai 1 ano |
|
softmax.cuh
|
ae1f211ce2
cuda : refactor into multiple files (#6269)
|
hai 1 ano |
|
sumrows.cu
|
ae1f211ce2
cuda : refactor into multiple files (#6269)
|
hai 1 ano |
|
sumrows.cuh
|
ae1f211ce2
cuda : refactor into multiple files (#6269)
|
hai 1 ano |
|
tsembd.cu
|
ae1f211ce2
cuda : refactor into multiple files (#6269)
|
hai 1 ano |
|
tsembd.cuh
|
ae1f211ce2
cuda : refactor into multiple files (#6269)
|
hai 1 ano |
|
unary.cu
|
f5ef34e428
feat: implemented sigmoid function (ggml/806)
|
hai 1 ano |
|
unary.cuh
|
f5ef34e428
feat: implemented sigmoid function (ggml/806)
|
hai 1 ano |
|
upscale.cu
|
48aa8fd1f2
ggml : add `ggml_upscale_ext` (ggml/814)
|
hai 1 ano |
|
upscale.cuh
|
ae1f211ce2
cuda : refactor into multiple files (#6269)
|
hai 1 ano |
|
vecdotq.cuh
|
e84b71c2c6
ggml : drop support for QK_K=64 (#7473)
|
hai 1 ano |
Georgi Gerganov
e84b71c2c6
ggml : drop support for QK_K=64 (#7473)