Fix LongRoPE KV Cache #1
Closed
Conversation
This file contains 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
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Outdated suggestions cannot be applied.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.
Issue
Before this fix: if we start from a prompt with length < 4K but the total sequence length will >4K after generation, the generated tokens after 4K will be gabage.
Root Cause
LongRoPE uses difference scaling factors for <=4k / > 4k sequences. When the generation comes across the switch point of 4K, the kv cache of prefill tokens are calculated based on short factors, but the new tokens are generated based on long factors. This inconsistency leads to the generation crash.
Solution
Pass
prompt_length + max_generation_tokensto the model.If it is larger than 4K, we should use the long factors for all calculations of this sequence.