Queue concurrency 1 to prevent CUDA OOM

Mandatory for me as I have a RTX 2070 (8Gb) and I get CUDA OOM if two users launch jobs at the same time. I can also use multiple tabs and jobs will be queued. You may not want it to be the default though.
2026-05-01 11:41:23 +00:00 · 2022-08-28 16:12:55 +02:00
parent 93e7dbaa71
commit 20c33f4423
1 changed files with 1 additions and 0 deletions
--- a/webui.py
+++ b/webui.py
@@ -1340,4 +1340,5 @@ demo = gr.TabbedInterface(
 """
 )

+demo.queue(concurrency_count=1)
 demo.launch()