memory : handle kv_unified for hybrid models #15050
Merged
Conversation
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
ggerganov
approved these changes
Aug 3, 2025
Nexesenex
pushed a commit
to Nexesenex/croco.cpp
that referenced
this pull request
Aug 5, 2025
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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Outdated suggestions cannot be applied.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.
Follow-up from #14725, which didn't really fix the underlying problem of not considering
cparams.kv_unified.Since #14959, inference with hybrid models has been broken (except when using
-kvu), due to hybrid memory not passingcparams.kv_unifiedproperly.Reproduction of the problem: attempt to run
llama-perplexitywith any hybrid model.$ ./bin/llama-perplexity -f /workspace/wikitext-2-raw/wiki.test.raw -m /workspace/gguf/LFM2-350M-BF16.gguf --chunks 10On
master, this fails with an assertionWith this PR, this is no longer a problem. I've tested this with https://huggingface.co/LiquidAI/LFM2-350M.
Make sure to read the contributing guidelines before submitting a PR