 Georgi Gerganov
|
183b3fac6c
metal : fix build errors and kernel sig after #2268 (#3898)
|
2 jaren geleden |
 cebtenzzre
|
898aeca90a
llama : implement YaRN RoPE scaling (#2268)
|
2 jaren geleden |
|
Georgi Gerganov
|
e16b9fa4ba
metal : multi-simd softmax (#3710)
|
2 jaren geleden |
|
Georgi Gerganov
|
71e3718abd
llama : refactor graph build code (#3837)
|
2 jaren geleden |
 Aarni Koskela
|
82a6646e02
metal : try cwd for ggml-metal.metal if bundle lookup fails (#3793)
|
2 jaren geleden |
|
Georgi Gerganov
|
469c9addef
metal : handle ggml_scale for n%4 != 0 (close #3754)
|
2 jaren geleden |
 Jhen-Jie Hong
|
c67fe68e41
metal : implement q5_0 and q5_1 kernels (#3648)
|
2 jaren geleden |
 Jan Ploski
|
f5f9121de1
llm : add MPT support (#3417)
|
2 jaren geleden |
|
Georgi Gerganov
|
fcca0a7004
refact : fix convert script + zero out KV cache to avoid nans (#3523)
|
2 jaren geleden |
|
Georgi Gerganov
|
dcc09d2596
metal : do not use mul_mm kernels when ne00 < 64 (#3542)
|
2 jaren geleden |
|
Georgi Gerganov
|
db3abcc114
sync : ggml (ggml-backend) (#3548)
|
2 jaren geleden |
|
Georgi Gerganov
|
94e502dfb7
ci : enable on obj-c changes + fix metal build (#3540)
|
2 jaren geleden |
|
Georgi Gerganov
|
b0ec5218c3
metal : support MTLGPUFamily < Apple7, formatting, style (#3524)
|
2 jaren geleden |
|
Jhen-Jie Hong
|
c26765a0a1
metal : support default.metallib load & reuse code for swift package (#3522)
|
2 jaren geleden |
 Phillip Kravtsov
|
0e797c2fc5
llm : support Adept Persimmon 8B (#3410)
|
2 jaren geleden |
 Jiahao Li
|
f56e1baec3
metal : alibi for arbitrary number of heads (#3426)
|
2 jaren geleden |
|
Georgi Gerganov
|
ec893798b7
llama : custom attention mask + parallel decoding + no context swaps (#3228)
|
2 jaren geleden |
 Rickard Hallerbäck
|
dc6897404e
metal : reusing llama.cpp logging (#3152)
|
2 jaren geleden |
|
Georgi Gerganov
|
8c00b7a6ff
sync : ggml (Metal F32 support + reduce ggml-alloc size) (#3192)
|
2 jaren geleden |
|
Georgi Gerganov
|
a51b687657
metal : relax conditions on fast matrix multiplication kernel (#3168)
|
2 jaren geleden |
 Kawrakow
|
f31b6f4e2d
metal : PP speedup (#3084)
|
2 jaren geleden |
 kchro3
|
21ac3a1503
metal : support for Swift (#3078)
|
2 jaren geleden |
|
Jhen-Jie Hong
|
4fd5477955
metal : support build for iOS/tvOS (#3089)
|
2 jaren geleden |
|
Kawrakow
|
be8c9c245b
metal : parallel RoPE on Metal (#3024)
|
2 jaren geleden |
 Przemysław Pawełczyk
|
fec2fb19e4
ggml : posixify madvise and pagesize (#3037)
|
2 jaren geleden |
|
Kawrakow
|
ca82cf7bac
metal : more optimizations (#2959)
|
2 jaren geleden |
 Karsten Weiss
|
8b56b4f2c3
metal : show all Metal device instances in the system (#2952)
|
2 jaren geleden |
|
Georgi Gerganov
|
13268c5331
metal : slight speed-up for add and mul kernels (#2917)
|
2 jaren geleden |
|
Kawrakow
|
e8d9158925
metal: somewhat faster f16 x f32 matrix multiply kernel (#2951)
|
2 jaren geleden |
|
Georgi Gerganov
|
3a007648f2
metal : add option to disable debug logs (close #2764)
|
2 jaren geleden |
