 Georgi Gerganov
|
2b3389677a
ggml : refactor rope norm/neox (#7634)
|
1 年之前 |
|
Georgi Gerganov
|
fb76ec31a9
ggml : fix YARN + add tests + add asserts (#7617)
|
1 年之前 |
|
Georgi Gerganov
|
0548a4187f
ggml : generalize GGML_OP_CONCAT (#7563)
|
1 年之前 |
|
Georgi Gerganov
|
1d8fca72ae
metal : add GGML_OP_REPEAT kernels (#7557)
|
1 年之前 |
|
Georgi Gerganov
|
62bfef5194
metal : disable FA kernel for HS=256 (#7556)
|
1 年之前 |
|
Georgi Gerganov
|
e84b71c2c6
ggml : drop support for QK_K=64 (#7473)
|
1 年之前 |
 liuwei-git
|
201cc11afa
llama : add phi3 128K model support (#7225)
|
1 年之前 |
|
Georgi Gerganov
|
6369bf0433
metal : handle F16 inf values, fix FA partial offload (#7434)
|
1 年之前 |
 John Balis
|
48aa8fd1f2
ggml : add `ggml_upscale_ext` (ggml/814)
|
1 年之前 |
|
Georgi Gerganov
|
e8a7fd4fb0
metal : support FA without mask + add asserts (#7278)
|
1 年之前 |
|
Georgi Gerganov
|
7bd4ffb780
metal : fix warnings (skipme) (#0)
|
1 年之前 |
 Justina Cho
|
f5ef34e428
feat: implemented sigmoid function (ggml/806)
|
1 年之前 |
|
Georgi Gerganov
|
9cb317f77e
ggml : full ALiBi support (#7192)
|
1 年之前 |
|
Georgi Gerganov
|
c0e6fbf8c3
metal : fix unused warning
|
1 年之前 |
|
Georgi Gerganov
|
9c67c2773d
ggml : add Flash Attention (#5021)
|
1 年之前 |
 slaren
|
0d56246f4b
ggml : group all experts in a single ggml_mul_mat_id (#6505)
|
1 年之前 |
 Shijie
|
f4dea7da18
llama : add qwen2moe (#6074)
|
1 年之前 |
 Dave
|
422c2aff1c
Added support for GGML_OP_CLAMP in Metal (#6662)
|
1 年之前 |
|
slaren
|
fbbc030ba9
metal : unify mul_mv_id kernels (#6556)
|
1 年之前 |
|
slaren
|
08a0c02060
ggml : mul_mat_id use the same tensor for all the experts (#6387)
|
1 年之前 |
 Kawrakow
|
cbc8343619
Make IQ1_M work for QK_K = 64 (#6327)
|
1 年之前 |
|
Kawrakow
|
55c1b2a3bb
IQ1_M: 1.75 bpw quantization (#6302)
|
1 年之前 |
|
Georgi Gerganov
|
b3e94f26ba
metal : proper assert for mat-mat memory alignment (#6225)
|
1 年之前 |
|
Kawrakow
|
76aa30a263
Add ability to use Q5_0, Q5_1, and IQ4_NL for quantized K cache (#6183)
|
1 年之前 |
|
Georgi Gerganov
|
381da2d9f0
metal : build metallib + fix embed path (#6015)
|
1 年之前 |
|
Georgi Gerganov
|
8030da7afe
ggml : reuse quantum structs across backends (#5943)
|
1 年之前 |
|
Kawrakow
|
44ca159faf
1.5 bit: we can do even better (#5999)
|
1 年之前 |
|
Kawrakow
|
be858f6205
Better 1.5 bit quantization (#5971)
|
1 年之前 |
|
Georgi Gerganov
|
bb6d00bbf9
metal : move mm_id indices to shared mem (#5982)
|
1 年之前 |
|
Georgi Gerganov
|
8a3012a4ad
ggml : add ggml-common.h to deduplicate shared code (#5940)
|
1 年之前 |
