Closed
Description
Environment
🐧 Linux
System
vllm/PyTorch 2.3.0+roc6.0, Firefox 126
Version
12.0 release & staging df2710 both affected
Desktop Information
No response
Describe the problem
Using a backend that supports multiple swipes per gen such as vllm
, if the user interrupts generation all but the first swipe are discarded.
Repro
- Set backend to
vllm
or something else with multiple swipes - Set swipes per gen to
8
- Send a message, receive 8 swipes
- Send a message and interrupt it, receive 1 swipe
Additional info
No response
Please tick the boxes
- I have explained the issue clearly, and I included all relevant info
- I have checked that this issue hasn't already been raised
- I have checked the docs