Klx/scaled softmax fix - #5061
Open
llaboon wants to merge 10 commits into
Open
Conversation
- Add diff operator with Triton kernel for kunlunxin backend
- Add square/square_ operators for kunlunxin backend
- Add string padding ('same'/'valid') support for conv3d
- Add cumprod/cumprod_ import to kunlunxin __init__.py
- Remove skipIf for conv1d_padding (Issue flagos-ai#3022 resolved)
- Remove skipIf for conv3d_padding (Issue flagos-ai#3023 resolved)
- Remove skip for softmax_backward dim=1 (Issue flagos-ai#2851 resolved)
Kunlunxin's version of transformer_engine (transformer_engine_klx) exposes a different signature for the GLU family primitives than upstream NVIDIA TE: - tex.geglu/reglu/swiglu require positional args (inp, fp8_meta_tensor, fp8_tensor, otype) instead of the (inp, quantizer=None) form the tests use - tex.dgeglu requires an otype argument; tex.dswiglu has no 'quantizer' kwarg In addition, the FlagGems Triton kernels for these ops currently fail to compile on XPU (xpu3-elfconv returns non-zero exit status 1), so the FlagGems side would need work as well. Skip these test files on Kunlunxin XPU. These ops are not registered via ATen dispatch, so no runtime crash is expected from user code - only the accuracy tests were failing.
…x-fix # Conflicts: # pyproject.toml # src/flag_gems/runtime/backend/_kunlunxin/ops/diff.py # src/flag_gems/runtime/backend/_kunlunxin/ops/softmax.py # src/flag_gems/runtime/backend/_kunlunxin/ops/square.py
llaboon
requested review from
0x45f,
103yiran,
Caeruleann,
bin913,
douxetpur,
hellojack163,
huangyiqun,
tengqm and
w1120029931-bit
as code owners
July 30, 2026 07:05
Signed-off-by: binliu <87918526+llaboon@users.noreply.github.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
PR Category
Type of Change
Description
Issue
Progress
Performance