Support grouped state heads in GDN attention - #21402
Conversation
Summary: Adds separate query-head and state-head configuration to the model definition as well as updates the fused channelwise_gated_delta_rule custom op to support grouped decode and prefill. Differential Revision: D113299906
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/21402
Note: Links to docs will display an error until the docs builds have been completed. ❌ 5 New Failures, 1 Unrelated FailureAs of commit c6b01e5 with merge base 82a63fd ( NEW FAILURES - The following jobs have failed:
FLAKY - The following job failed but was likely due to flakiness present on trunk:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
|
@JakeStevens has exported this pull request. If you are a Meta employee, you can view the originating Diff in D113299906. |
This PR needs a
|
Summary:
Adds separate query-head and state-head configuration to the model definition as well as updates the fused
channelwise_gated_delta_rule custom op to support grouped decode and prefill.
Differential Revision: D113299906