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

llama : add check for KV cache shifts (#10401)

Browse source 
ggml-ci
This commit is contained in:
Georgi Gerganov 2024-11-19 13:29:26 +02:00 • committed by GitHub
parent a88ad007de
commit 8e752a777b
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
3 changed files with 14 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

6
common/common.cpp
View file

@ -875,6 +875,12 @@ struct common_init_result common_init_from_params(common_params & params) {
return iparams;
}
if (params.ctx_shift && !llama_kv_cache_can_shift(lctx)) {
LOG_ERR("%s: KV cache shifting is not supported for this model (--no-context-shift to disable)'\n", __func__);
llama_free_model(model);
return iparams;
}
if (!params.control_vectors.empty()) {
if (params.control_vector_layer_start <= 0) params.control_vector_layer_start = 1;
if (params.control_vector_layer_end <= 0) params.control_vector_layer_end = llama_n_layer(model);