Improve prefix handling, attention mask compatibility, and KV cache control#11
Open
YangYangGirl wants to merge 2 commits intoQwenLM:mainfrom
Open
Improve prefix handling, attention mask compatibility, and KV cache control#11YangYangGirl wants to merge 2 commits intoQwenLM:mainfrom
YangYangGirl wants to merge 2 commits intoQwenLM:mainfrom
Conversation
- Prefix mask fix Changed prefix mask values from 0 to 1 so virtual prefix tokens can properly participate in attention. - Prefix K/V concatenation during training Added logic to concatenate prefix key/value states during training. - 2D attention mask support Enabled attention_mask in [B, T] format in addition to 4D [B, H, T_q, T_k]. - Configurable KV cache usage Added use_cache parameter; KV cache is disabled during training.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Changed prefix mask values from 0 to 1 so virtual prefix tokens can properly participate in attention.
Added logic to concatenate prefix key/value states during training.
Enabled attention_mask in [B, T] format in addition to 4D [B, H, T_q, T_k].
Added use_cache parameter; KV cache is disabled during training.