 Jeff Bolz
|
886b97a5d6
tests: Generate unique input values for count_equal (#15487)
|
hace 5 meses |
|
Jeff Bolz
|
c9a24fb932
vulkan: Support FA with any multiple of 8 head sizes (#15537)
|
hace 5 meses |
|
Jeff Bolz
|
611f419cff
vulkan: optimize rms_norm, and allow the work to spread across multiple SMs (#15281)
|
hace 5 meses |
 Acly
|
0a9b43e507
vulkan : support ggml_mean (#15393)
|
hace 5 meses |
 rmatif
|
92f7f0a53c
ggml: add `conv3d` op (#15182)
|
hace 5 meses |
|
Jeff Bolz
|
96452a3fa4
vulkan: Reuse conversion results in prealloc_y (#15410)
|
hace 5 meses |
|
Jeff Bolz
|
de5627910d
vulkan: Optimize argsort (#15354)
|
hace 5 meses |
|
Jeff Bolz
|
1fe00296f5
vulkan: fuse adds (#15252)
|
hace 5 meses |
|
Jeff Bolz
|
2e2b22ba66
vulkan: Add missing bounds checking to scalar/coopmat1 mul_mat_id (#15334)
|
hace 5 meses |
 Georgi Gerganov
|
5edf1592fd
vulkan : fix out-of-bounds access in argmax kernel (#15342)
|
hace 5 meses |
 Jonathan Graehl
|
5cdb27e091
finetune: SGD optimizer, more CLI args (#13873)
|
hace 5 meses |
 Oliver Simons
|
6028bf7435
CUDA: Optimize `reduce_rows_f32` kernel, leading up to 25x perf improvement on kernel-level and 10% perf increase for Gemma3n (#15132)
|
hace 5 meses |
|
Georgi Gerganov
|
fd1234cb46
llama : add gpt-oss (#15091)
|
hace 5 meses |
|
Jeff Bolz
|
ec0b18802c
vulkan: Support ne[3]>1 in noncontig matrix-vector multiply (#15015)
|
hace 5 meses |
 Sigbjørn Skjæret
|
138b288b59
cuda : add softcap fusion (#14907)
|
hace 5 meses |
 Leonard Mosescu
|
bda62193b2
test-backend-ops : extend test case filtering (#14865)
|
hace 5 meses |
 Erik Scholz
|
89d1029559
vulkan : add fp16 support for the conv_2d kernel (#14872)
|
hace 6 meses |
 Aman Gupta
|
446595b9b3
Docs: add instructions for adding backends (#14889)
|
hace 6 meses |
|
Georgi Gerganov
|
18f3b5ff9e
tests : add non-cont K,V FA tests
|
hace 6 meses |
|
Aman Gupta
|
8c988fa41d
CUDA: add fused rms norm (#14800)
|
hace 6 meses |
|
Jeff Bolz
|
c2e058f1b4
vulkan/cuda: Fix im2col when KW!=KH (#14789)
|
hace 6 meses |
 Ervin Áron Tasnádi
|
a979ca22db
ggml: adds CONV_2D op and direct GEMM Vulkan implementation (#14316)
|
hace 6 meses |
|
Georgi Gerganov
|
bf9087f59a
metal : fuse add, mul + add tests (#14596)
|
hace 6 meses |
|
Georgi Gerganov
|
225e7a1438
llama : add high-throughput mode (#14363)
|
hace 6 meses |
 Tarek Dakhran
|
c31e60647d
tests : cover lfm2 cases in test_ssm_conv (#14651)
|
hace 6 meses |
|
Acly
|
3e303b1107
vulkan : implement ggml_roll (ggml/1290)
|
hace 6 meses |
|
Aman Gupta
|
11ee0fea2a
Docs: script to auto-generate ggml operations docs (#14598)
|
hace 6 meses |
 compilade
|
a57d1bcb3c
cuda : support Falcon-H1 state size for SSM_SCAN (#14602)
|
hace 6 meses |
 Xuan-Son Nguyen
|
98bab638fb
ggml : add ggml_scale_bias (#14417)
|
hace 6 meses |
|
Georgi Gerganov
|
4d0dcd4a06
cuda : fix rope with partial rotation and non-cont src (#14580)
|
hace 6 meses |
