ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-01-26 17:20:01 +00:00
Code Issues Packages Projects Releases Wiki Activity

Arghh, we need to increase the context size again

Browse Source
This commit is contained in:
Kawrakow
2026-01-12 14:59:15 +02:00
parent 9821ac7b9c
commit 74dc8aa99e
1 changed files with 4 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
src/llama-load-tensors.cpp
View File

@@ -199,7 +199,10 @@ create_tensors_helper::create_tensors_helper(llama_model_loader & _ml, llama_mod
buft_layer_count[model.buft_layer[i].buft_matrix]++;
}
ctx_size = ggml_tensor_overhead()*(ml.n_tensors + 1); // +1 for models where tok_embd is duplicated as output
auto n_tensors = ml.n_tensors;
if (ml.merge_qkv) n_tensors += n_layer;
if (ml.merge_up_gate_exps) n_tensors += n_layer;
ctx_size = ggml_tensor_overhead()*(n_tensors + 1); // +1 for models where tok_embd is duplicated as output
ctx_size += ggml_tensor_overhead()*n_layer*3; // for moe merged tensors
if (model.splits.size() > 1) {