Better CPU FA performance for DeepSeek-Lite #410
Merged
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Outdated suggestions cannot be applied.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.
This FA tweak improves DeepSeek-Lite CPU TG performance with
Q8_0KV cache.Not sure if it will have a positive impact for the large DeepSeek models. To optimize the FA strategy for those I need to be able to test, which I cannot atm.
The graph shows a comparison between the main branch and this PR for a
Q4_0quantized DeepSeek-Lite model. The CPU is Ryzen-7950X. The x-axis isN_KV/1000, whereN__KVis the number of tokens in the K cache, which is quantized withQ8_0. Thesweep-benchcommand wasMain branch
PR