We are rolling out a fix that will restore any lingering reduction in prefill throughput
We have deployed a fix and capacity has recovered
The new Tinker SDK version 0.23.3 has been pushed. Please update your SDK if you are running into this issue.
This issue has been resolved
The issue is fully resolved now.
All models should be back up now!
All training should be running normally
API has recovered. Root cause is identified and we will push that fix out soon.
We've pushed a fix for this issue.
·