mirror of
https://github.com/kvcache-ai/ktransformers.git
synced 2026-11-05 01:07:28 +00:00
[GH-ISSUE #11] Is Flash Attention 2 Necessary for Qwen2Moe? #5
Labels
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference
kvcache-ai/ktransformers#5
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Originally created by @cherrymorning on GitHub (Jul 29, 2024).
Original GitHub issue: https://github.com/kvcache-ai/ktransformers/issues/11
I tried to run Qwen2Moe but I only have access to T4 GPUs, which don't support Flash Attention 2.
Is Flash Attention 2 a strict requirement for Qwen2Moe?
Any guidance would be greatly appreciated!
@Atream commented on GitHub (Jul 29, 2024):
Flash Attention 2 is required for Qwen2_moe, or it may overflow. This is Qwen2_moe's own character.
probability tensor contains eitherinf,nanor element < 0failed. #327