Add quantized KV cache for MLX draft#89
Open
Ziqiao-git wants to merge 1 commit intoz-lab:mainfrom
Open
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Adds
quantize_kv_bits=4/8toload_draft()and--draft-quantize-kv-bitsto benchmark CLI.Similar to the existing
sliding_window_sizeoption, this bounds the draft KV cache memory footprint (int4 = 4x smaller, int8 = 2x smaller).Implementation:
DFlashAttentiondequantizes cache output before concatenating with fp16 proposal KV. This keeps the cache compact without changing the SDPA kernel.Spot-checked on Qwen3.5-4B (gsm8k, thinking mode): outputs identical across fp16/int4/int8 for the first 500 chars. Acceptance within noise. Happy to add unit tests if needed.