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

handle seq rm return value

Browse source
This commit is contained in:
Jan Boon 2024-03-30 02:39:33 +08:00
parent bf1d4932f8
commit 8ab1a17251
2 changed files with 7 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

5
examples/server/server.cpp
View file

@ -1745,7 +1745,10 @@ struct server_context {
// Erase token cache // Erase token cache
const size_t n_erased = slot->cache_tokens.size(); const size_t n_erased = slot->cache_tokens.size();
llama_kv_cache_seq_rm(ctx, slot->id + 1, -1, -1); if (!llama_kv_cache_seq_rm(ctx, slot->id + 1, -1, -1)) {
send_error(task, "Failed to erase slot KV cache", ERROR_TYPE_INVALID_REQUEST);
break;
}
slot->cache_tokens.clear(); slot->cache_tokens.clear();
server_task_result result; server_task_result result;

4
llama.cpp
View file

@ -15243,7 +15243,9 @@ size_t llama_state_seq_set_data(struct llama_context * ctx, const uint8_t * src,
GGML_ASSERT(!kv_self.recurrent); // not implemented GGML_ASSERT(!kv_self.recurrent); // not implemented
// Wipe the slot // Wipe the slot
llama_kv_cache_seq_rm(kv_self, dest_seq_id, -1, -1); if (!llama_kv_cache_seq_rm(kv_self, dest_seq_id, -1, -1)) {
return 0;
}
const uint8_t * inp = src; const uint8_t * inp = src;