Lumpiasty/llama.cpp
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

server : support slot save/restore with media inputs (#26640)

Browse Source 
* server : save serialized image chunks at the end of the llama state

* server : support multimodal slot state save/restore with packed payload

* server : refine image slot state serialization

* server : support media slot state and centralize media validation

* server : remove unnecessary comment

* server : remove defensive media checks and move the chunk type check to validate()
This commit is contained in:
Chipmunk
2026-08-12 12:20:28 +02:00
committed by GitHub
parent 4dd127584b
commit 5d9e5ac30e
7 changed files with 593 additions and 78 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/llama-context.cpp
+11
View File
@@ -3110,6 +3110,17 @@ size_t llama_context::state_seq_load_file(llama_seq_id seq_id, const char * file
{
const uint32_t n_token_count = file.read_u32();
if (tokens_out == nullptr) {
const size_t n_token_max = (file.size() - file.tell()) / sizeof(llama_token);
if (n_token_count > n_token_max) {
LLAMA_LOG_ERROR("%s: token count in sequence state file exceeds the file size! %u > %zu\n", __func__, n_token_count, n_token_max);
return 0;
}
*n_token_count_out = n_token_count;
return file.tell();
}
if (n_token_count > n_token_capacity) {
LLAMA_LOG_ERROR("%s: token count in sequence state file exceeded capacity! %u > %zu\n", __func__, n_token_count, n_token_capacity);
return 0;