millaguie/llama.cpp
1
0 Fork 0
Code Issues Pull Requests Actions 28 Packages Projects Releases Wiki Activity

llama-batch: add n_keep_tail in split_equal for recurrent models (#25278)

Browse Source
This commit is contained in:
Aman Gupta authored and GitHub committed 2026-07-08 15:55:19 +08:00
1 parent f296fdfbed
commit 230ea9d214
10 files changed
+102 -38
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

src/llama-kv-cache-dsv4.cpp
+1 -1
View File
@@ -1110,7 +1110,7 @@ llama_memory_context_ptr llama_kv_cache_dsv4::init_batch(
if (has_coupled) {
ubatch = balloc.split_seq(n_ubatch);
} else {
ubatch = balloc.split_equal(n_ubatch, raw_per_seq || comp_per_seq);
ubatch = balloc.split_equal(n_ubatch, raw_per_seq || comp_per_seq, 0);
}
if (ubatch.n_tokens == 0) {