Disable FlashAttenion for is_causal=True when seqlen q not equal kv #111007
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
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/111007
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit be9bb0e with merge base 652f4c6 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
drisspg
force-pushed
the
disable_flash_on_seqlen_q_not_equal_kv
branch
from
5dfadec to
044a820
Compare
drisspg
added
module: multi-headed-attention
topic: not user facing
cpuhrsch
approved these changes
Oct 19, 2023
drisspg
force-pushed
the
disable_flash_on_seqlen_q_not_equal_kv
branch
from
e9f3845 to
be9bb0e
Compare
|
@pytorchbot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
Merge failedReason: 1 mandatory check(s) failed. The first few are: Dig deeper by viewing the failures on hud |
|
@pytorchbot merge -f "unrelated failures" |
|
@pytorchbot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
pytorchmergebot
pushed a commit
that referenced
this pull request
Oct 25, 2023
# Summary We were restricted from updating to the newest version of FlashAttention based off of the changes to is_casual described here: #108108 Prior to this PR we landed: #111007 which enabled us to updated beyond: 9e5e8bc91e on FlashAttentionV2. With this PR we have updated to this commit: Dao-AILab/flash-attention@02ac572. Or Tag 2.3.2 ## Plans Following this PR I plan to work more on #110681 in order to expose a CausalVariant attn_mask, w/ the potential for also exposing a kvcache attn_mask. Pull Request resolved: #111886 Approved by: https://github.com/cpuhrsch
xuhancn
pushed a commit
to xuhancn/pytorch
that referenced
this pull request
Nov 7, 2023
…ytorch#111007) # Summary: This pull request **removes** support for non-square sequence lengths in causal attention when using FlashAttention V2. ### Why are doing this // FlashAttention 2 updated the default mask meaning for causal in this PR: // 9e5e8bc91e it is now aligned to lower_right which would be a BC break // for non-square masks. We will not support non-square masks for causal w/ FAV2 For more context see: pytorch#108108 ### Followup A large number of people will likely want to use FAV2 with lower_right causal attention for non equal sequence lengths. See this RFC : pytorch#110681 Pull Request resolved: pytorch#111007 Approved by: https://github.com/cpuhrsch
xuhancn
pushed a commit
to xuhancn/pytorch
that referenced
this pull request
Nov 7, 2023
# Summary We were restricted from updating to the newest version of FlashAttention based off of the changes to is_casual described here: pytorch#108108 Prior to this PR we landed: pytorch#111007 which enabled us to updated beyond: 9e5e8bc91e on FlashAttentionV2. With this PR we have updated to this commit: Dao-AILab/flash-attention@02ac572. Or Tag 2.3.2 ## Plans Following this PR I plan to work more on pytorch#110681 in order to expose a CausalVariant attn_mask, w/ the potential for also exposing a kvcache attn_mask. Pull Request resolved: pytorch#111886 Approved by: https://github.com/cpuhrsch
Skylion007
pushed a commit
to Skylion007/pytorch
that referenced
this pull request
Nov 14, 2023
…ytorch#111007) # Summary: This pull request **removes** support for non-square sequence lengths in causal attention when using FlashAttention V2. ### Why are doing this // FlashAttention 2 updated the default mask meaning for causal in this PR: // 9e5e8bc91e it is now aligned to lower_right which would be a BC break // for non-square masks. We will not support non-square masks for causal w/ FAV2 For more context see: pytorch#108108 ### Followup A large number of people will likely want to use FAV2 with lower_right causal attention for non equal sequence lengths. See this RFC : pytorch#110681 Pull Request resolved: pytorch#111007 Approved by: https://github.com/cpuhrsch
Skylion007
pushed a commit
to Skylion007/pytorch
that referenced
this pull request
Nov 14, 2023
# Summary We were restricted from updating to the newest version of FlashAttention based off of the changes to is_casual described here: pytorch#108108 Prior to this PR we landed: pytorch#111007 which enabled us to updated beyond: 9e5e8bc91e on FlashAttentionV2. With this PR we have updated to this commit: Dao-AILab/flash-attention@02ac572. Or Tag 2.3.2 ## Plans Following this PR I plan to work more on pytorch#110681 in order to expose a CausalVariant attn_mask, w/ the potential for also exposing a kvcache attn_mask. Pull Request resolved: pytorch#111886 Approved by: https://github.com/cpuhrsch
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 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.
Summary:
This pull request removes support for non-square sequence lengths in causal attention when using FlashAttention V2.
Why are doing this
// FlashAttention 2 updated the default mask meaning for causal in this PR:
// 9e5e8bc91e it is now aligned to lower_right which would be a BC break
// for non-square masks. We will not support non-square masks for causal w/ FAV2
For more context see:
#108108
Followup
A large number of people will likely want to use FAV2 with lower_right causal attention for non equal sequence lengths. See this RFC : #110681