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

keep in the size check

Browse source
This commit is contained in:
Jan Boon 2024-03-30 02:49:31 +08:00
parent 0d2213678c
commit 29f18c29b4
1 changed files with 6 additions and 6 deletions
Download patch file Download diff file Expand all files Collapse all files

12
llama.cpp
View file

@ -15343,13 +15343,13 @@ size_t llama_state_seq_set_data(struct llama_context * ctx, const uint8_t * src,
memcpy(&v_size_el_ref, inp, sizeof(v_size_el_ref)); memcpy(&v_size_el_ref, inp, sizeof(v_size_el_ref));
inp += sizeof(v_size_el_ref); inp += sizeof(v_size_el_ref);
if (cell_count) { const size_t v_size_el = ggml_type_size(kv_self.v_l[il]->type);
const size_t v_size_el = ggml_type_size(kv_self.v_l[il]->type); if (v_size_el != v_size_el_ref) {
if (v_size_el != v_size_el_ref) { llama_kv_cache_seq_rm(kv_self, dest_seq_id, -1, -1);
llama_kv_cache_seq_rm(kv_self, dest_seq_id, -1, -1); return 0;
return 0; }
}
if (cell_count) {
// For each row in the transposed matrix, read the values for the whole cell range // For each row in the transposed matrix, read the values for the whole cell range
for (uint32_t j = 0; j < n_embd_v_gqa; ++j) { for (uint32_t j = 0; j < n_embd_v_gqa; ++j) {
const size_t dst_offset = (kv_head + j * kv_size) * v_size_el; const size_t dst_offset = (kv_head + j * kv_size) * v_size_el;