diff options
| author | Mitja Felicijan <mitja.felicijan@gmail.com> | 2026-02-12 20:57:17 +0100 |
|---|---|---|
| committer | Mitja Felicijan <mitja.felicijan@gmail.com> | 2026-02-12 20:57:17 +0100 |
| commit | b333b06772c89d96aacb5490d6a219fba7c09cc6 (patch) | |
| tree | 211df60083a5946baa2ed61d33d8121b7e251b06 /llama.cpp/ggml/src/ggml-cuda/softmax.cuh | |
| download | llmnpc-b333b06772c89d96aacb5490d6a219fba7c09cc6.tar.gz | |
Engage!
Diffstat (limited to 'llama.cpp/ggml/src/ggml-cuda/softmax.cuh')
| -rw-r--r-- | llama.cpp/ggml/src/ggml-cuda/softmax.cuh | 7 |
1 files changed, 7 insertions, 0 deletions
diff --git a/llama.cpp/ggml/src/ggml-cuda/softmax.cuh b/llama.cpp/ggml/src/ggml-cuda/softmax.cuh new file mode 100644 index 0000000..93dfee8 --- /dev/null +++ b/llama.cpp/ggml/src/ggml-cuda/softmax.cuh @@ -0,0 +1,7 @@ +#include "common.cuh" + +#define CUDA_SOFT_MAX_BLOCK_SIZE 1024 + +void ggml_cuda_op_soft_max(ggml_backend_cuda_context & ctx, ggml_tensor * dst); + +void ggml_cuda_op_soft_max_back(ggml_backend_cuda_context & ctx, ggml_tensor * dst); |