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Outdated suggestions cannot be applied.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.
The current version Flash Attention does not support the ALiBi positional embedding. This PR adds support for the Triton-based Flash Attention that does allow a bias added to the self-attention calculation.
This can be enabled simply by setting
"pos_emb": "alibi"
and adding"flash"
to the attention config in your model's configuration.To support this feature we bumped up the listed Triton dependency to
triton==2.0.0.dev20221202
and in the course of implementing it we discovered that SparseAttention became broken at some point in the past, see #863.