 Jeff Bolz
|
a0374a67e2
vulkan: Handle updated FA dim2/3 definition (#14518)
|
6 bulan lalu |
 Sigbjørn Skjæret
|
ddef99522d
server : fix assistant prefilling when content is an array (#14360)
|
6 bulan lalu |
|
Sigbjørn Skjæret
|
6681688146
opencl: add GELU_ERF (#14476)
|
6 bulan lalu |
 Georgi Gerganov
|
bac8bed248
eval-callback : check for empty input (#14539)
|
6 bulan lalu |
 R0CKSTAR
|
b81510a7b7
test-backend-ops: add support for specifying output format (#14368)
|
6 bulan lalu |
|
Georgi Gerganov
|
ef797db357
metal : disable fast math in all quantize kernels (#14528)
|
6 bulan lalu |
|
Georgi Gerganov
|
67d1ef23c6
batch : add optional for sequential equal split (#14511)
|
6 bulan lalu |
|
Georgi Gerganov
|
7b50f7c025
graph : prepare for 4D mask (#14515)
|
6 bulan lalu |
|
Georgi Gerganov
|
c79184d2d1
batch : add n_used count (#14512)
|
6 bulan lalu |
 luyhcsu
|
499a8f5a78
CANN: Replace aclrtMemsetSync with aclnnInplaceZero operator (#14002)
|
6 bulan lalu |
|
Sigbjørn Skjæret
|
28657a8229
ggml : implement GEGLU_ERF and GEGLU_QUICK ops (#14445)
|
6 bulan lalu |
 lhez
|
bee28421be
opencl : broadcast for soft_max (#14510)
|
6 bulan lalu |
|
Jeff Bolz
|
2b72bedec1
vulkan: support mixed/deepseekR1 FA head sizes (#14509)
|
6 bulan lalu |
 Johannes Gäßler
|
c8c4495b8d
ggml: backward pass for split swiglu (#14483)
|
6 bulan lalu |
 Nicolò Scipione
|
7b63a71a6b
Fix conditional enabling following arch checks for ggml-sycl (#14504)
|
6 bulan lalu |
 Xuan-Son Nguyen
|
0c2ee38ab7
convert : correct gemma 3n conversion (#14450)
|
6 bulan lalu |
|
Georgi Gerganov
|
a70c8a0c4b
kv-cache : use ggml_set_rows (#14285)
|
6 bulan lalu |
|
Georgi Gerganov
|
9067487c44
ggml : fix FA mask dim 2 and 3 (#14505)
|
6 bulan lalu |
|
Georgi Gerganov
|
d4cdd9c1c3
ggml : remove kompute backend (#14501)
|
6 bulan lalu |
 Aman Gupta
|
55c2646b45
CUDA: add dynamic shared mem to softmax, refactor general usage (#14497)
|
6 bulan lalu |
|
Sigbjørn Skjæret
|
e75ba4c043
gguf-py : add support for chat template jinja files (#14508)
|
6 bulan lalu |
 compilade
|
5d46babdc2
llama : initial Mamba-2 support (#9126)
|
6 bulan lalu |
|
Georgi Gerganov
|
e17991c466
sync : ggml
|
6 bulan lalu |
 Daniel Bevenius
|
c46944aa25
ggml : add version function to get lib version (ggml/1286)
|
6 bulan lalu |
 Rotem Dan
|
f3ed38d793
Set RPATH to "@loader_path" / "$ORIGIN" to ensure executables and dynamic libraries search for dependencies in their origin directory. (#14309)
|
6 bulan lalu |
|
Aman Gupta
|
55a1c5a5fd
CUDA: add softmax broadcast (#14475)
|
6 bulan lalu |
|
Johannes Gäßler
|
12a81af45f
CUDA: broadcasting for FlashAttention mask (#14500)
|
6 bulan lalu |
|
Jeff Bolz
|
8875523eb3
vulkan: support softmax/FA batch and broadcast (#14449)
|
6 bulan lalu |
|
Georgi Gerganov
|
ec68e84c32
ggml : support bcast ggml_soft_max_ext, ggml_flash_attn_ext (#14435)
|
6 bulan lalu |
 zhouwg
|
307e79d33d
opencl : fix possible buffer overflow in dump_tensor (#14490)
|
6 bulan lalu |
