 Johannes Gäßler
|
e11bd856d5
CPU/CUDA: Gemma 2 FlashAttention support (#8542)
|
1 year ago |
 compilade
|
a1631e53f6
llama : simplify Mamba with advanced batch splits (#8526)
|
1 year ago |
 Daniel Bevenius
|
06943a69f6
ggml : move rope type enum to ggml.h (#8949)
|
1 year ago |
 Molly Sophia
|
2d5dd7bb3f
ggml : add epsilon as a parameter for group_norm (#8818)
|
1 year ago |
|
Daniel Bevenius
|
655858ace0
ggml : move c parameter comment to ggml_rope_ext (ggml/901)
|
1 year ago |
 Sigbjørn Skjæret
|
b72c20b85c
Fix conversion of unnormalized BF16->BF16 weights (#7843)
|
1 year ago |
 slaren
|
2b1f616b20
ggml : reduce hash table reset cost (#8698)
|
1 year ago |
 Georgi Gerganov
|
eddcb5238b
ggml : add and use ggml_cpu_has_llamafile() (#8664)
|
1 year ago |
 hipudding
|
1bdd8ae19f
[CANN] Add Ascend NPU backend (#6035)
|
1 year ago |
|
Georgi Gerganov
|
370b1f7e7a
ggml : minor naming changes (#8433)
|
1 year ago |
 Dibakar Gope
|
0f1a39f343
ggml : add AArch64 optimized GEMV and GEMM Q4 kernels (#5780)
|
1 year ago |
|
Georgi Gerganov
|
f3f65429c4
llama : reorganize source code + improve CMake (#8006)
|
1 year ago |
