| .. |
|
template-instances
|
69c487f4ed
CUDA: MMQ code deduplication + iquant support (#8495)
|
hai 1 ano |
|
vendors
|
c35e586ea5
musa: enable building fat binaries, enable unified memory, and disable Flash Attention on QY1 (MTT S80) (#9526)
|
hai 1 ano |
|
acc.cu
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
acc.cuh
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
arange.cu
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
arange.cuh
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
argmax.cu
|
fabdc3bda3
ggml/ex: calculate accuracy in graph, adapt MNIST (ggml/980)
|
hai 1 ano |
|
argmax.cuh
|
fabdc3bda3
ggml/ex: calculate accuracy in graph, adapt MNIST (ggml/980)
|
hai 1 ano |
|
argsort.cu
|
2b1f616b20
ggml : reduce hash table reset cost (#8698)
|
hai 1 ano |
|
argsort.cuh
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
binbcast.cu
|
424c5d00a9
ggml/examples: add backend support for numerical optimization (ggml/949)
|
hai 1 ano |
|
binbcast.cuh
|
424c5d00a9
ggml/examples: add backend support for numerical optimization (ggml/949)
|
hai 1 ano |
|
clamp.cu
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
clamp.cuh
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
common.cuh
|
fabdc3bda3
ggml/ex: calculate accuracy in graph, adapt MNIST (ggml/980)
|
hai 1 ano |
|
concat.cu
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
concat.cuh
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
conv-transpose-1d.cu
|
fde13b3bb9
feat: cuda implementation for `ggml_conv_transpose_1d` (ggml/854)
|
hai 1 ano |
|
conv-transpose-1d.cuh
|
fde13b3bb9
feat: cuda implementation for `ggml_conv_transpose_1d` (ggml/854)
|
hai 1 ano |
|
convert.cu
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
convert.cuh
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
count-equal.cu
|
fabdc3bda3
ggml/ex: calculate accuracy in graph, adapt MNIST (ggml/980)
|
hai 1 ano |
|
count-equal.cuh
|
fabdc3bda3
ggml/ex: calculate accuracy in graph, adapt MNIST (ggml/980)
|
hai 1 ano |
|
cpy.cu
|
116efee0ee
cuda: add q8_0->f32 cpy operation (#9571)
|
hai 1 ano |
|
cpy.cuh
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
cross-entropy-loss.cu
|
424c5d00a9
ggml/examples: add backend support for numerical optimization (ggml/949)
|
hai 1 ano |
|
cross-entropy-loss.cuh
|
424c5d00a9
ggml/examples: add backend support for numerical optimization (ggml/949)
|
hai 1 ano |
|
dequantize.cuh
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
diagmask.cu
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
diagmask.cuh
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
dmmv.cu
|
7a11eb3a26
cuda : fix dmmv cols requirement to 2*GGML_CUDA_DMMV_X (#8800)
|
hai 1 ano |
|
dmmv.cuh
|
7a11eb3a26
cuda : fix dmmv cols requirement to 2*GGML_CUDA_DMMV_X (#8800)
|
hai 1 ano |
|
fattn-common.cuh
|
e11bd856d5
CPU/CUDA: Gemma 2 FlashAttention support (#8542)
|
hai 1 ano |
|
fattn-tile-f16.cu
|
fabdc3bda3
ggml/ex: calculate accuracy in graph, adapt MNIST (ggml/980)
|
hai 1 ano |
|
fattn-tile-f16.cuh
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
fattn-tile-f32.cu
|
c35e586ea5
musa: enable building fat binaries, enable unified memory, and disable Flash Attention on QY1 (MTT S80) (#9526)
|
hai 1 ano |
|
fattn-tile-f32.cuh
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
fattn-vec-f16.cuh
|
fabdc3bda3
ggml/ex: calculate accuracy in graph, adapt MNIST (ggml/980)
|
hai 1 ano |
|
fattn-vec-f32.cuh
|
e11bd856d5
CPU/CUDA: Gemma 2 FlashAttention support (#8542)
|
hai 1 ano |
|
fattn-wmma-f16.cuh
|
e11bd856d5
CPU/CUDA: Gemma 2 FlashAttention support (#8542)
|
hai 1 ano |
|
fattn.cu
|
a5b57b08ce
CUDA: enable Gemma FA for HIP/Pascal (#9581)
|
hai 1 ano |
|
fattn.cuh
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
getrows.cu
|
2b1f616b20
ggml : reduce hash table reset cost (#8698)
|
hai 1 ano |
|
getrows.cuh
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
im2col.cu
|
aaa4099925
CUDA: remove bad assert (ggml/972)
|
hai 1 ano |
|
im2col.cuh
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
mma.cuh
|
808aba3916
CUDA: optimize and refactor MMQ (#8416)
|
hai 1 ano |
|
mmq.cu
|
5af118efda
CUDA: fix --split-mode row race condition (#9413)
|
hai 1 ano |
|
mmq.cuh
|
5af118efda
CUDA: fix --split-mode row race condition (#9413)
|
hai 1 ano |
|
mmvq.cu
|
2b1f616b20
ggml : reduce hash table reset cost (#8698)
|
hai 1 ano |
|
mmvq.cuh
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
norm.cu
|
2d5dd7bb3f
ggml : add epsilon as a parameter for group_norm (#8818)
|
hai 1 ano |
|
norm.cuh
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
opt-step-adamw.cu
|
424c5d00a9
ggml/examples: add backend support for numerical optimization (ggml/949)
|
hai 1 ano |
|
opt-step-adamw.cuh
|
424c5d00a9
ggml/examples: add backend support for numerical optimization (ggml/949)
|
hai 1 ano |
|
out-prod.cu
|
d13edb17ed
ggml : fix builds (#0)
|
hai 1 ano |
|
out-prod.cuh
|
424c5d00a9
ggml/examples: add backend support for numerical optimization (ggml/949)
|
hai 1 ano |
|
pad.cu
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
pad.cuh
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
pool2d.cu
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
pool2d.cuh
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
quantize.cu
|
2b1f616b20
ggml : reduce hash table reset cost (#8698)
|
hai 1 ano |
|
quantize.cuh
|
808aba3916
CUDA: optimize and refactor MMQ (#8416)
|
hai 1 ano |
|
rope.cu
|
06943a69f6
ggml : move rope type enum to ggml.h (#8949)
|
hai 1 ano |
|
rope.cuh
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
rwkv-wkv.cu
|
2a63caaa69
RWKV v6: RWKV_WKV op CUDA implementation (#9454)
|
hai 1 ano |
|
rwkv-wkv.cuh
|
2a63caaa69
RWKV v6: RWKV_WKV op CUDA implementation (#9454)
|
hai 1 ano |
|
scale.cu
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
scale.cuh
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
softmax.cu
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
softmax.cuh
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
sum.cu
|
5cb12f6839
CUDA: fix sum.cu compilation for CUDA < 11.7 (#9562)
|
hai 1 ano |
|
sum.cuh
|
202084d31d
tests: add gradient tests for all backends (ggml/932)
|
hai 1 ano |
|
sumrows.cu
|
231cff5f6f
sync : ggml
|
hai 1 ano |
|
sumrows.cuh
|
231cff5f6f
sync : ggml
|
hai 1 ano |
|
tsembd.cu
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
tsembd.cuh
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
unary.cu
|
2a63caaa69
RWKV v6: RWKV_WKV op CUDA implementation (#9454)
|
hai 1 ano |
|
unary.cuh
|
2a63caaa69
RWKV v6: RWKV_WKV op CUDA implementation (#9454)
|
hai 1 ano |
|
upscale.cu
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
upscale.cuh
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
hai 1 ano |
|
vecdotq.cuh
|
69c487f4ed
CUDA: MMQ code deduplication + iquant support (#8495)
|
hai 1 ano |
Johannes Gäßler
fabdc3bda3
ggml/ex: calculate accuracy in graph, adapt MNIST (ggml/980)