Repository navigation
Conversation
naamfung
added a commit
to naamfung/laamaafung
that referenced
this pull request
Jul 14, 2026
Checkpoints are session-internal KV state snapshots for hybrid/recurrent model rollback. They are meaningless across sessions - a new session creates its own. Previously alloc() copied the source prompt's checkpoints into each cache entry, causing checkpoint RAM to accumulate within the cache-ram budget without benefit. Store empty checkpoints in cache entries instead. When load() restores a cached prompt, the new session correctly starts with no checkpoints. Ref: ggml-org/llama.cpp#24649
naamfung
added a commit
to naamfung/laamaafung
that referenced
this pull request
Jul 14, 2026
Checkpoints are session-internal KV state snapshots for hybrid/recurrent model rollback. They are meaningless across sessions - a new session creates its own. Previously alloc() copied the source prompt's checkpoints into each cache entry, causing checkpoint RAM to accumulate within the cache-ram budget without benefit. Store empty checkpoints in cache entries instead. When load() restores a cached prompt, the new session correctly starts with no checkpoints. Ref: ggml-org/llama.cpp#24649
naamfung
added a commit
to naamfung/laamaafung
that referenced
this pull request
Jul 20, 2026
Checkpoints are session-internal KV state snapshots for hybrid/recurrent model rollback. They are meaningless across sessions - a new session creates its own. Previously alloc() copied the source prompt's checkpoints into each cache entry, causing checkpoint RAM to accumulate within the cache-ram budget without benefit. Store empty checkpoints in cache entries instead. When load() restores a cached prompt, the new session correctly starts with no checkpoints. Ref: ggml-org/llama.cpp#24649
This was referenced Aug 16, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
…ent OOM
Checkpoints are session-internal rollback points for hybrid/recurrent models (e.g. Qwen3.6) and are not needed for inter-session cache reuse. Previously, prompt_save() copied all checkpoint data into the cache entry, and alloc() copied checkpoints into each new cache state, causing checkpoint RAM to accumulate up to the cache-ram limit (default 8 GiB) on top of the active slot checkpoint budget.
Fix: clear checkpoint data in prompt_save() before serialising the KV state, and store empty checkpoints in new cache entries in alloc(). Also adds a Linux-only get_available_ram_bytes() helper to log available RAM at save time for easier diagnosis.
Overview
Additional information
Requirements