Skip to content

[PyTorch] GDN2 support and linear attention refactor #3521

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

Open
ksivaman wants to merge 11 commits into
base: main
Choose a base branch
from
Open

[PyTorch] GDN2 support and linear attention refactor #3521

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
11 commits
Select commit Hold shift + click to select a range
5c0e47e
GDN2 attention
ksivaman Sep 11, 2026
f9bb3ae
Update cudnnFE requirement
ksivaman Sep 11, 2026
3566b27
Merge branch 'NVIDIA:main' into gdn2_attention
ksivaman Sep 15, 2026
de902ac
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot] Sep 15, 2026
7aa3b61
Minor doc changes and review
ksivaman Sep 15, 2026
ab2cf2f
Remove negative test cases
ksivaman Sep 15, 2026
5e1dc0f
Merge branch 'main' into gdn2_attention
ksivaman Sep 16, 2026
b506d72
Remove fp8 autocast checks
ksivaman Sep 16, 2026
2495c5f
Merge branch 'main' into gdn2_attention
ksivaman Sep 16, 2026
ae7676a
review
ksivaman Sep 17, 2026
65e5133
docsting
ksivaman Sep 17, 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
2 changes: 1 addition & 1 deletion build_tools/pytorch.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -31,7 +31,7 @@ def install_requirements() -> List[str]:
"packaging",
"pydantic",
"nvdlfw-inspect",
"nvidia-cudnn-frontend>=1.28.0",
"nvidia-cudnn-frontend>=1.29.0",

@greptile-apps greptile-apps Bot Sep 15, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P1 Required runtime is not installed

GDN2 explicitly requires the cuDNN frontend cutedsl runtime, but both installation declarations request only the base nvidia-cudnn-frontend package. If that base package does not include the optional cutlass runtime, the newly mandatory L0 GDN2 test fails during collection and the advertised feature cannot run.

]


Expand Down
3 changes: 3 additions & 0 deletions docs/api/pytorch.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -28,6 +28,9 @@ PyTorch
.. autoapiclass:: transformer_engine.pytorch.GatedDeltaNetAttention(num_attention_heads, kv_channels, **kwargs)
:members: forward

.. autoapiclass:: transformer_engine.pytorch.GatedDeltaNet2Attention(num_attention_heads, kv_channels, **kwargs)
:members: forward

.. autoapiclass:: transformer_engine.pytorch.MultiheadAttention(hidden_size, num_attention_heads, **kwargs)
:members: forward, set_context_parallel_group, set_tensor_parallel_group

Expand Down
2 changes: 1 addition & 1 deletion pyproject.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,7 +3,7 @@
# See LICENSE for license information.

[build-system]
requires = ["setuptools>=61.0", "cmake>=3.21", "wheel", "pybind11[global]", "ninja", "pip", "torch>=2.1", "jax>=0.5.0", "flax>=0.7.1", "nvidia-cudnn-frontend>=1.28.0"]
requires = ["setuptools>=61.0", "cmake>=3.21", "wheel", "pybind11[global]", "ninja", "pip", "torch>=2.1", "jax>=0.5.0", "flax>=0.7.1", "nvidia-cudnn-frontend>=1.29.0"]

# Use legacy backend to import local packages in setup.py
build-backend = "setuptools.build_meta:__legacy__"
1 change: 1 addition & 0 deletions qa/L0_pytorch_unittest/test.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -64,6 +64,7 @@ python3 -m pytest --tb=auto --junitxml=$XML_LOG_DIR/pytest_test_identity_quantiz
NVTE_ALLOW_UNSAFE_PICKLE_EXTRA_STATE=1 python3 -m pytest --tb=auto --junitxml=$XML_LOG_DIR/pytest_test_attention.xml $TE_PATH/tests/pytorch/attention/test_attention.py || test_fail "test_attention.py"
python3 -m pytest --tb=auto --junitxml=$XML_LOG_DIR/pytest_test_flex_attention.xml $TE_PATH/tests/pytorch/attention/test_flex_attention.py || test_fail "test_flex_attention.py"
NVTE_GDN_TEST_REQUIRED=1 python3 -m pytest --tb=auto --junitxml=$XML_LOG_DIR/pytest_test_gdn_attention.xml $TE_PATH/tests/pytorch/attention/test_gdn_attention.py || test_fail "test_gdn_attention.py"
NVTE_GDN2_TEST_REQUIRED=1 python3 -m pytest --tb=auto --junitxml=$XML_LOG_DIR/pytest_test_gdn2_attention.xml $TE_PATH/tests/pytorch/attention/test_gdn2_attention.py || test_fail "test_gdn2_attention.py"
NVTE_ALLOW_UNSAFE_PICKLE_EXTRA_STATE=1 NVTE_ALLOW_NONDETERMINISTIC_ALGO=0 python3 -m pytest --tb=auto --junitxml=$XML_LOG_DIR/pytest_test_attention_deterministic.xml $TE_PATH/tests/pytorch/attention/test_attention.py || test_fail "NVTE_ALLOW_NONDETERMINISTIC_ALGO=0 test_attention.py"
python3 -m pytest --tb=auto --junitxml=$XML_LOG_DIR/pytest_test_linear_mxfp8_attention.xml $TE_PATH/tests/pytorch/attention/test_linear_mxfp8_attention.py || test_fail "test_linear_mxfp8_attention.py"
python3 -m pytest --tb=auto --junitxml=$XML_LOG_DIR/pytest_test_fused_mla_q_uproj.xml $TE_PATH/tests/pytorch/attention/test_fused_mla_q_uproj.py || test_fail "test_fused_mla_q_uproj.py"
Expand Down
Loading
Loading