[GH-ISSUE #171] Add an option to disable flash attention for older GPU's #134

Open
opened 2026-10-26 21:12:51 +00:00 by bsakaguchi · 2 comments
Owner

Originally created by @apcameron on GitHub (Feb 14, 2025).
Original GitHub issue: https://github.com/deepseek-ai/Janus/issues/171

When I run generation_inference.py I get the error below.

RuntimeError: FlashAttention only supports Ampere GPUs or newer.

Please add an option to either disable it.

Originally created by @apcameron on GitHub (Feb 14, 2025). Original GitHub issue: https://github.com/deepseek-ai/Janus/issues/171 When I run generation_inference.py I get the error below. RuntimeError: FlashAttention only supports Ampere GPUs or newer. Please add an option to either disable it.
Author
Owner

@shootheart commented on GitHub (May 19, 2025):

Same problem. Is there a way to disable it manually?

<!-- gh-comment-id:2889510525 --> @shootheart commented on GitHub (May 19, 2025): Same problem. Is there a way to disable it manually?
Author
Owner

@hongjx175 commented on GitHub (May 19, 2025):

A simple way is to modify the config.json

<!-- gh-comment-id:2889732227 --> @hongjx175 commented on GitHub (May 19, 2025): A simple way is to modify the `config.json`
Sign in to join this conversation.
No labels
pull-request
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
deepseek-ai/Janus#134
No description provided.