ver4a/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

RoPE: fix back, CUDA support for back + noncont. (#11240)

Browse source 
* RoPE: fix back, CUDA support for back + noncont.

* fix comments reg. non-cont. RoPE support [no-ci]
This commit is contained in:
Johannes Gäßler 2025-01-15 12:51:37 +01:00 • committed by GitHub
parent 0ccd7f3eb2
commit 432df2d5f9
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
9 changed files with 269 additions and 258 deletions
Download patch file Download diff file Expand all files Collapse all files

2
ggml/src/ggml-cuda/rope.cuh
View file

@ -3,3 +3,5 @@
#define CUDA_ROPE_BLOCK_SIZE 256
void ggml_cuda_op_rope(ggml_backend_cuda_context & ctx, ggml_tensor * dst);
void ggml_cuda_op_rope_back(ggml_backend_cuda_context & ctx, ggml_tensor * dst);