happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Fix byte count for 3d or higher tensors

This commit is contained in:
Ed Addario 2025-08-22 09:01:57 +01:00
parent ec0afbe79f
commit 35c1504441
No known key found for this signature in database
GPG Key ID: E7875815A3230993
1 changed files with 3 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
src/llama-quant.cpp
View File

@ -676,10 +676,9 @@ static std::unordered_map<std::string, ggml_type> target_bpw_type(
auto tensor_bytes = [](const ggml_tensor * t, const ggml_type typ) -> size_t {
const int64_t n_per_row = t->ne[0];
const int64_t nrows = t->ne[1];
const int64_t ne2 = t->ne[2] > 0 ? t->ne[2] : 1;
const size_t row_sz = ggml_row_size(typ, n_per_row);
return (size_t)ne2 * (size_t)nrows * row_sz;
const size_t row_sz = ggml_row_size(typ, n_per_row);
const int64_t nrows = ggml_nrows(t);
return (size_t)nrows * row_sz;
};
auto tensor_bpw = [&](const ggml_tensor * t, const ggml_type typ) -> double {