Skip to content

Releases: bdashore3/flash-attention

v2.7.1.post1

06 Dec 06:09
Compare
Choose a tag to compare

v2.7.0.post2

03 Dec 19:23
Compare
Choose a tag to compare
Actions: Bump Cuda

12.4

Signed-off-by: kingbri <[email protected]>

v2.6.3

26 Jul 00:14
Compare
Choose a tag to compare

Synced to Upstream version

NOTE: Backward and Dropout are disabled meaning that this release is INFERENCE ONLY.

This is because including these features more than doubles the build time and makes the github action time itself out. Please raise an issue to the parent repo to help reduce the build times if you want these features.

v2.6.1

12 Jul 00:06
Compare
Choose a tag to compare
Actions: Switch to CUDA 12.3

Signed-off-by: kingbri <[email protected]>

v2.5.9.post2

09 Jul 23:28
Compare
Choose a tag to compare
v2.5.9.post2 Pre-release
Pre-release

Quick release to add softcapping commits. Does not have backward, dropout, or alibi support.

v2.5.9.post1

28 May 01:45
Compare
Choose a tag to compare
Actions: Clarify dispatch formatting

Signed-off-by: kingbri <[email protected]>

v2.5.8

28 Apr 07:55
Compare
Choose a tag to compare

Same as Upstream tag

Now built for only torch 2.2.2 and 2.3.0

v2.5.6

30 Mar 20:34
Compare
Choose a tag to compare

v2.5.2

07 Feb 22:37
Compare
Choose a tag to compare

Same as the upstream tag

Adds this PR to help fix building on Windows

v2.4.2

03 Feb 00:29
Compare
Choose a tag to compare

Inline with the parent repo's tag

Made for cuda 12.x and pytorch 2.1.2 and 2.2

v2.4.3 and up cannot be built on Windows at this time.