]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
overview / pkg / ggml / sources / llama.cpp / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: b8eeb87)
memory : fix broken batch splits for recurrent cache (#14575)
authorcompilade <redacted>
Tue, 8 Jul 2025 15:37:47 +0000 (11:37 -0400)
committerGitHub <redacted>
Tue, 8 Jul 2025 15:37:47 +0000 (18:37 +0300)
commitbb4f7a9e4eec171fecf0f640b1337a1c24485560
treeb63a6cc4b8b2cc4b954580001a2e5b20624e4d8ftree
parentb8eeb8741d4483daf576498cf90537b4de71633ccommit | diff
memory : fix broken batch splits for recurrent cache (#14575)

Splits producing more than one ubatch per batch for recurrent models
were broken with #14512.

This fixes it by moving the completeness check after the ubatch split loop.
src/llama-memory-recurrent.cpp diff | blob | history
Packaging of ggml-org/llama.cpp