Skip to content

Issues: triton-lang/triton

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

696 Open 922 Closed
696 Open 922 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Assignee
Filter by who’s assigned
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Issues list

tl.dot output is different from torch.mm bug
#5849 opened Feb 7, 2025 by rozyangno
tl.dot silent failure when used on 3D tensors bug needs reproducer
#5842 opened Feb 6, 2025 by mandroid6
2
Gather does not work if index is much longer than value bug
#5836 opened Feb 6, 2025 by SchrodingerZhu
@Mogball
Potential Bug in **_attn_fwd_tma** Function bug
#5816 opened Feb 5, 2025 by qinxiangyujiayou
Triton interpreter cannot handle parameters that alias bug
#5791 opened Feb 2, 2025 by saagarjha
13
"ImportError: cannot import name 'backends' from 'triton.backends' (unknown location)" for triton installed from source bug
#5773 opened Jan 31, 2025 by martin-kukla
Misreport "Cannot have return statements inside while or for" if values returned by function are disregarded bug
#5768 opened Jan 30, 2025 by xinyazhang
[3.2.x] ptx_get_version cannot handle CUDA>12.6
#5737 opened Jan 29, 2025 by h-vetinari
@bertmaher
8
No module named triton.backends.nvidia._C when installing from triton-3.2.0-cp312-cp312-linux_x86_64.whl bug
#5728 opened Jan 28, 2025 by Steboss
[RFC] Improve performance for layer-norm in turtorial performance
#5712 opened Jan 27, 2025 by ywq880611
1
The Precision Issue of the GELU Operator question
#5692 opened Jan 24, 2025 by noc-turne
3
[Dialect] Implement Proton device buffer init and alloc ops
#5689 opened Jan 24, 2025 by CRobeck
Implement Proton version of the global scratch alloc
#5688 opened Jan 24, 2025 by CRobeck
FusedAttention Bug (at backward with causal=False) bug
#5660 opened Jan 21, 2025 by SanggeunParrk
BarrierOp in AssertOp lowering can cause deadlock bug
#5632 opened Jan 17, 2025 by kishore-ganesh
5
tl.dot with batched (3D) input is working only in emulation mode when os.environ['TRITON_INTERPRET'] = '1' is set. bug
#5620 opened Jan 15, 2025 by warpuv
IR builder does not support FP8 dtypes bug
#5601 opened Jan 14, 2025 by danielvegamyhre
FP8 GEMM implemented with triton is slower on Ada (SM89) performance
#5583 opened Jan 12, 2025 by TechxGenus
FMAs now using a lot more registers than necessary (and often spilling) performance
#5572 opened Jan 10, 2025 by vwbaker
1
No wheels for arm bug
#5561 opened Jan 9, 2025 by surak
No tag / release for 3.1 bug
#5560 opened Jan 9, 2025 by surak
Triton 3.1.0 failed with a simple tl.dot and then tl.store example bug
#5557 opened Jan 9, 2025 by annyan09023
5
"Triton is 4x faster than the Torch JIT" in tutorial 2 seems outdated.
#5536 opened Jan 6, 2025 by lifengxiang1025
2
compile error build
#5535 opened Jan 6, 2025 by JaneConan
2
max(tensor, axis = 1) doesn't work without TRITON_INTERPRET=1 bug needs reproducer
#5522 opened Jan 3, 2025 by wa008
3
ProTip! Updated in the last three days: updated:>2025-02-04.