RL icon indicating copy to clipboard operation
RL copied to clipboard
verified publisher icon NVIDIA-NeMo

v0.5 Performance Improvement on Long Max-Seqlen

Open guyueh1 opened this issue 1 month ago • 2 comments

Tracking items to improve performance for long max-seqlen workloads, including training and generation performance at long context.

guyueh1 avatar Nov 07 '25 00:11 guyueh1

@youngeunkwon0405 you can create sub-issues once you identified concrete items; If they are general items not async specific you can assign them to me.

guyueh1 avatar Nov 07 '25 00:11 guyueh1

Sure I will do.

youngeunkwon0405 avatar Nov 07 '25 01:11 youngeunkwon0405

There is no items planned for v0.5 actually. We will plan for v0.6.

guyueh1 avatar Dec 02 '25 21:12 guyueh1

Just for the logging purpose, I will share some knowledge I have found so far. https://docs.google.com/presentation/d/1U-Gd6gMmH6QX3KPzqwcyAFbK5Uir7XiDXe5j59EsP1s/edit?usp=sharing

youngeunkwon0405 avatar Dec 02 '25 22:12 youngeunkwon0405

@guyueh1 you moved this one to no action necessary. is there anything remaining to do on this issue?

terrykong avatar Dec 04 '25 06:12 terrykong

@terrykong No features or fixes will be ready for v0.5, we will plan for v0.6 if that's alright

guyueh1 avatar Dec 10 '25 21:12 guyueh1