mirror of
https://github.com/deepseek-ai/DeepGEMM.git
synced 2026-11-05 00:47:26 +00:00
[GH-ISSUE #222] [QUESTION] Can fp8_paged_mqa_logits support next_n > 2? #127
Labels
No labels
pull-request
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference
deepseek-ai/DeepGEMM#127
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Originally created by @starwang1024 on GitHub (Oct 24, 2025).
Original GitHub issue: https://github.com/deepseek-ai/DeepGEMM/issues/222
I noticed that the code here in
fp8_paged_mqa_logitsonly supports next_n <= 2. May I ask what the reason for this is? Additionally, what modifications would be needed to support next_n > 2?Any reply would be very helpful.
@LyricZhao commented on GitHub (Dec 5, 2025):
Internally at DeepSeek, we don't use
next_n > 2. But the NV https://github.com/deepseek-ai/DeepGEMM/tree/nv_dev branch supports such features.@Paiiiiiiiiiiiiii commented on GitHub (Dec 7, 2025):
I noticed that there is one line (DG_HOST_ASSERT(next_n == 1 or next_n == 2 or next_n == 4);), So now it still dont support nextn=3?
@aiyxxj commented on GitHub (Dec 31, 2025):
I'd like to know if there are any plans to support nextn>2 on the main branch.
@hpz4311 commented on GitHub (Apr 13, 2026):
Hello, has this issue been finally resolved? I directly removed the assertion for next_n, and I found that running the test script works fine with next_n = 3, next_n = 4, and so on. Is it okay to just remove the assertion?