tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-08-18 05:56:00 -04:00
Code Issues Packages Projects Releases Wiki Activity

CUDA: noncont MMVQ + batched bs1 MUL_MAT_ID (#13014)

Browse Source 
* CUDA: noncont MMVQ + batched bs1 MUL_MAT_ID

* fix logic for RoPE support, CUDA graphs
This commit is contained in:
Johannes Gäßler
2025-04-22 21:27:40 +02:00
committed by GitHub
parent dc39a5e7a8
commit 658987cfc9
9 changed files with 548 additions and 426 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
ggml/src/ggml-cuda/mmvq.cuh
View File

@@ -2,6 +2,9 @@
#define MMVQ_MAX_BATCH_SIZE 8 // Max. batch size for which to use MMVQ kernels.
void ggml_cuda_mul_mat_vec_q(ggml_backend_cuda_context & ctx,
const ggml_tensor * src0, const ggml_tensor * src1, const ggml_tensor * ids, ggml_tensor * dst);
void ggml_cuda_op_mul_mat_vec_q(
ggml_backend_cuda_context & ctx,
const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst, const char * src0_dd_i, const float * src1_ddf_i,