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

cuda : fix rope + add tests (#7452)

Browse source 
* cuda : fix rope pos data

ggml-ci

* ggml : drop mode & 1 == 1 support for ggml_rope

ggml-ci

* ggml : support freq_factors for f16 rope (CPU)

ggml-ci

* tests : add rope tests using frequency factors

ggml-ci
This commit is contained in:
Georgi Gerganov 2024-05-22 11:01:35 +03:00 • committed by GitHub
parent 201cc11afa
commit 3e5faa8503
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
4 changed files with 47 additions and 20 deletions
Download patch file Download diff file Expand all files Collapse all files

4
ggml-cuda/rope.cu
View file

@ -283,9 +283,9 @@ void ggml_cuda_op_rope(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
const bool is_neox = mode & 2;
const bool is_glm = mode & 4;
if (is_neox) {
pos = (const int32_t *) src1_d;
pos = (const int32_t *) src1_d;
if (is_neox) {
if (src2 != nullptr) {
freq_factors = (const float *) src2->data;
}