char-fork/llama.cpp

git clone https://git.t4t.associates/char-fork/llama.cpp

Aman GuptaCUDA: fuse adds, fuse add with rms norm (#15631)009b709d6

char/aliceai-t5-moe
brancheschar/aliceai-t5-moeall refs →
838 B18 linesraw
1#include "common.cuh"
2
3void ggml_cuda_op_norm(ggml_backend_cuda_context & ctx, ggml_tensor * dst);
4
5void ggml_cuda_op_group_norm(ggml_backend_cuda_context & ctx, ggml_tensor * dst);
6
7void ggml_cuda_op_rms_norm(ggml_backend_cuda_context & ctx, ggml_tensor * dst);
8
9void ggml_cuda_op_rms_norm_fused(ggml_backend_cuda_context & ctx, ggml_tensor * dst, ggml_tensor * mul_tensor);
10
11void ggml_cuda_op_rms_norm_fused_add(ggml_backend_cuda_context & ctx,
12                                     ggml_tensor *               dst,
13                                     ggml_tensor *               mul_tensor,
14                                     ggml_tensor *               add_tensor);
15
16void ggml_cuda_op_rms_norm_back(ggml_backend_cuda_context & ctx, ggml_tensor * dst);
17
18void ggml_cuda_op_l2_norm(ggml_backend_cuda_context & ctx, ggml_tensor * dst);