Skip to content

fix(rocm): reuse staged moe alignment on Triton 3.8 #691

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
jason-fxz merged 2 commits into from
Oct 15, 2026
Merged

fix(rocm): reuse staged moe alignment on Triton 3.8 #691

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
f2a7a72
fix(rocm): reuse staged moe alignment below triton 3.9
LZ-QWQ Oct 14, 2026
ef32685
fix(rocm): restrict moe alignment workaround to Triton 3.8
LZ-QWQ Oct 15, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
13 changes: 12 additions & 1 deletion python/freetoken/moe/fused.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -84,7 +84,18 @@ def moe_align_block_size(
- The padding ensures that the total number of tokens is now divisible
by block_size for proper block matrix operations.
"""
from freetoken.kernel.backend import is_sgl_kernel_installed
from freetoken.kernel.backend import is_rocm, is_sgl_kernel_installed

if is_rocm():
import triton
from packaging.version import Version

# FIXME: remove once Triton 3.8 is unsupported. Its AMD backend folds compares on tl.histogram results to false.
# https://github.com/triton-lang/triton/pull/11246
if Version(triton.__version__).release[:2] == (3, 8):
from freetoken.kernel import moe_align_block_size_triton

return moe_align_block_size_triton(topk_ids, block_size, num_experts)

if not is_sgl_kernel_installed():
from freetoken.kernel.triton.moe_align import (
Expand Down
Loading