llama : offload KV cache per-layer
This commit is contained in:
parent
c294c78eb7
commit
986b3da76a
2 changed files with 86 additions and 151 deletions
2
llama.h
2
llama.h
|
@ -196,6 +196,8 @@ extern "C" {
|
|||
bool f16_kv; // use fp16 for KV cache, fp32 otherwise
|
||||
bool logits_all; // the llama_eval() call computes all logits, not just the last one
|
||||
bool embedding; // embedding mode only
|
||||
bool offload_k;
|
||||
bool offload_v;
|
||||
};
|
||||
|
||||
// model quantization parameters
|
||||
|
|
Loading…
Add table
Add a link
Reference in a new issue