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

remove next_metadata_size

Browse source
This commit is contained in:
ngxson 2024-03-28 00:18:11 +01:00
parent 83f944c22d
commit 38d0e1da9a
1 changed files with 2 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

7
examples/gguf-split/gguf-split.cpp
View file

@ -226,13 +226,10 @@ struct split_strategy {
size_t curr_tensors_size = 0; // current size by counting only tensors size (without metadata)
for (int i = 0; i < n_tensors; ++i) {
struct ggml_tensor * t = ggml_get_tensor(ctx_meta, gguf_get_tensor_name(ctx_gguf, i));
// calculate the "imaginary" size including this tensor
// calculate the "imaginary" size = the current size + next tensor size
size_t n_bytes = GGML_PAD(ggml_nbytes(t), GGUF_DEFAULT_ALIGNMENT);
size_t next_tensors_size = curr_tensors_size + n_bytes;
size_t next_metadata_size = gguf_get_meta_size(ctx_out)
+ GGUF_DEFAULT_ALIGNMENT
+ sizeof(struct ggml_tensor *); // not sure if we're missing something
if (should_split(i, next_tensors_size + next_metadata_size)) {
if (should_split(i, next_tensors_size)) {
new_ctx_out();
curr_tensors_size = n_bytes;
} else {