Iwan Kawrakow 2a5552830b MoE improvements on Metal ...

This version beats mainline, there are things I don't understand:
* Mianline has effectively gone to GEMV for MUL_MAT_ID. We can do the
  same, but we are 30% slower. Why?
* Using actual GEMM, we beat mainline with ubtach size of 128. But then
  performance degrades. Why?

2025-04-02 15:26:19 +02:00

..

cmake

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

include

Use bf16 instead of fp16 block scales for q8_1 (#292)

2025-03-27 05:49:16 +01:00

src

MoE improvements on Metal

2025-04-02 15:26:19 +02:00

.gitignore

Merge mainline llama.cpp (#3)

2024-07-27 07:55:01 +02:00

CMakeLists.txt

Compile time option to use bf16 for qunts without MMQ kernels (#261)

2025-03-18 07:37:10 +01:00