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

DeepseekV4: fix rollback with multi-seq (#26756)

Browse Source 
* DeepseekV4: fix rollback with multi-seq

* fix model loading

* make pending rollback single use

* only clear cache for seq_id for full load

* add assert for compress ratio

* make graph topology static

* pass true instead of flags in clear_compressed

* cont : clean-up + TODOs

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
This commit is contained in:
Aman Gupta
2026-08-23 13:57:49 +03:00
committed by GitHub
co-authored by Georgi Gerganov
parent d3371929bb
commit b0539c43ed
10 changed files with 345 additions and 62 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/llama-context.cpp
-4
View File
@@ -3218,8 +3218,6 @@ size_t llama_context::state_read_data(llama_io_read_i & io) {
}
size_t llama_context::state_seq_write_data(llama_io_write_i & io, llama_seq_id seq_id, llama_state_seq_flags flags) {
GGML_UNUSED(seq_id);
if (memory) {
memory->state_write(io, seq_id, flags);
}
@@ -3228,8 +3226,6 @@ size_t llama_context::state_seq_write_data(llama_io_write_i & io, llama_seq_id s
}
size_t llama_context::state_seq_read_data(llama_io_read_i & io, llama_seq_id seq_id, llama_state_seq_flags flags) {
GGML_UNUSED(seq_id);
if (memory) {
memory->state_read(io, seq_id, flags);
}