Commits · 2f11bd2e7b56befc2c3e5994d427e5fe885a2ae3 · OpenDAS / TransformerEngine

26 Aug, 2025 2 commits
- [DCU] fix · 4927d10e
  yuguo authored Aug 26, 2025
  
  4927d10e
- [DCU] fix · 2e870ed9
  yuguo authored Aug 26, 2025
  
  2e870ed9
25 Aug, 2025 1 commit
- [DCU] fix moe tensorwise int8 · 059d92e2
  yuguo authored Aug 25, 2025
  
  059d92e2
23 Aug, 2025 2 commits
- [DCU] tensorwise int8 gemm surpport bias · 11864d3d
  yuguo authored Aug 23, 2025
  
  11864d3d
- [DCU] fix tensorwise int8 moe bugs · 32edae18
  yuguo authored Aug 23, 2025
  
  32edae18
21 Aug, 2025 2 commits
- fix · 0cf10d1c
  yuguo authored Aug 21, 2025
  
  0cf10d1c
- [DCU] tensorwise int8 train opt · 7a923605
  yuguo authored Aug 21, 2025
  
  7a923605
19 Aug, 2025 1 commit
- add swap env · d19a5a44
  evt_fugx1 authored Aug 19, 2025
  
  d19a5a44
13 Aug, 2025 2 commits
- Formatted · c6dae0e5
  wenjh authored Aug 13, 2025
```
Signed-off-by: wenjh <wenjh@sugon.com>
```
  c6dae0e5
- Delete tmpArgs in groupedgemm · 3cdceb87
  wenjh authored Aug 13, 2025
```
Signed-off-by: wenjh <wenjh@sugon.com>
```
  3cdceb87
12 Aug, 2025 1 commit
- Use hipMemcpy rather than hipMemcpyAsync · 1f97aebb
  wenjh authored Aug 12, 2025
```
Signed-off-by: wenjh <wenjh@sugon.com>
```
  1f97aebb
11 Aug, 2025 1 commit
- Fix hipblaslt_goupedgemm · 8665c111
  wenjh authored Aug 11, 2025
```
Signed-off-by: wenjh <wenjh@sugon.com>
```
  8665c111
08 Aug, 2025 2 commits
- [DCU] add tmp_userArgs for blaslt groupgemm · 14aaa48a
  yuguo authored Aug 08, 2025
  
  14aaa48a
- [DCU] add NVTE_INT8_SIM_FP8_TENSORWISE · 3eb6ea62
  yuguo authored Aug 08, 2025
  
  3eb6ea62
07 Aug, 2025 1 commit
- [DCU] fix channelwise train accumulate bug · 68d6c506
  yuguo authored Aug 07, 2025
  
  68d6c506
06 Aug, 2025 3 commits
- [DCU] fix channelwise train oom bug · 4a013bd5
  yuguo authored Aug 06, 2025
  
  4a013bd5
- [DCU] userargs groupedgemm · ddfbdaf4
  yuguo authored Aug 06, 2025
  
  ddfbdaf4
- [DCU] fix batchlinear core dump in 2.5 · eac0d49b
  yuguo authored Aug 06, 2025
  
  eac0d49b
05 Aug, 2025 1 commit
- [DCU] fix all gather usage · d8041744
  yuguo authored Aug 05, 2025
  
  d8041744
25 Jul, 2025 2 commits
- Disable lightop for now · ff1e14af
  wenjh authored Jul 25, 2025
```
Signed-off-by: wenjh <wenjh@sugon.com>
```
  ff1e14af
- TN w8a8 lightop · 4088797f
  wenjh authored Jul 25, 2025
```
Signed-off-by: wenjh <wenjh@sugon.com>
```
  4088797f
22 Jul, 2025 2 commits
- [DCU] fix compile · 782f6092
  yuguo authored Jul 22, 2025
  
  782f6092
- [DCU] fix merge · eac75188
  yuguo authored Jul 22, 2025
  
  eac75188
18 Jul, 2025 4 commits

[DCU] fix · d9847b6d
yuguo authored Jul 18, 2025

d9847b6d
[DCU] fix bias gradient · 736e8f8b
yuguo authored Jul 18, 2025

736e8f8b
Optional lightop · f5349823
wenjh authored Jul 18, 2025
```
Signed-off-by: wenjh <wenjh@sugon.com>
```
f5349823

Update cudnn-frontend to 1.13.0 (#1960) · 657c965b

Charlene Yang authored Jul 17, 2025



* update cudnn-frontend to 1.13.0
Signed-off-by: Charlene Yang <8636796+cyanguwa@users.noreply.github.com>

* disable 9.11 for a bug
Signed-off-by: Charlene Yang <8636796+cyanguwa@users.noreply.github.com>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci



* fix selection logic
Signed-off-by: Charlene Yang <8636796+cyanguwa@users.noreply.github.com>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci



---------
Signed-off-by: Charlene Yang <8636796+cyanguwa@users.noreply.github.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

657c965b

17 Jul, 2025 6 commits

[Common] Optimize KV cache related kernels (#1914) · f8933bbf

Charlene Yang authored Jul 17, 2025



* optimize kv_cache reindex and copy kernels
Signed-off-by: Charlene Yang <8636796+cyanguwa@users.noreply.github.com>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci



* avoid reindexing from python side
Signed-off-by: Charlene Yang <8636796+cyanguwa@users.noreply.github.com>

* rename variable from previous commit
Signed-off-by: Charlene Yang <8636796+cyanguwa@users.noreply.github.com>

* minor fix
Signed-off-by: Charlene Yang <8636796+cyanguwa@users.noreply.github.com>

* minor fix
Signed-off-by: Charlene Yang <8636796+cyanguwa@users.noreply.github.com>

---------
Signed-off-by: Charlene Yang <8636796+cyanguwa@users.noreply.github.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

f8933bbf

[JAX] Remove unneccessary MXFP8 scale_inv padding (#1954) · 5350f277

Phuong Nguyen authored Jul 17, 2025



* remove unnecessary padding
Signed-off-by: Phuong Nguyen <phuonguyen@nvidia.com>

* adapt the test_distributed_layernorm byte count
Signed-off-by: Phuong Nguyen <phuonguyen@nvidia.com>


---------
Signed-off-by: Phuong Nguyen <phuonguyen@nvidia.com>

5350f277

Use lightop replace w8a8_mutmal_extension · 1036ccfe
wenjh authored Jul 17, 2025
```
Signed-off-by: wenjh <wenjh@sugon.com>
```
1036ccfe
[DCU] fix · 00738a42
yuguo authored Jul 17, 2025

00738a42

[PyTorch] Add save_original_input in Linear/GroupedLinear to save memory (#1865) · 07afda98

hx authored Jul 17, 2025



* save original input
Signed-off-by: Hongxiao Bai <hongxiaob@nvidia.com>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci



* fix input_quantizer usage in Linear bwd
Signed-off-by: Hongxiao Bai <hongxiaob@nvidia.com>

* minor fix
Signed-off-by: Hongxiao Bai <hongxiaob@nvidia.com>

* refine the docstring
Signed-off-by: Hongxiao Bai <hongxiaob@nvidia.com>

* Merge remote-tracking branch 'origin/main' into save_bf16_in_fp8_gemm
Signed-off-by: Hongxiao Bai <hongxiaob@nvidia.com>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci



* decouple linear bwd with save_original_input; clean up UTs
Signed-off-by: Hongxiao Bai <hongxiaob@nvidia.com>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci



---------
Signed-off-by: Hongxiao Bai <hongxiaob@nvidia.com>
Signed-off-by: hx <hongxiaob@nvidia.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Xin Yao <xiny@nvidia.com>
Co-authored-by: Tim Moon <4406448+timmoon10@users.noreply.github.com>

07afda98

mxfp8 (for all gemm layouts) is not supported on 120+ arch yet (#1939) · fa91ed72

Sudhakar Singh authored Jul 16, 2025



* mxfp8 is not supported on 120+ arch yet
Signed-off-by: Sudhakar Singh <sudhakars@nvidia.com>

* change the default recipe for arch 120
Signed-off-by: Sudhakar Singh <sudhakars@nvidia.com>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci



---------
Signed-off-by: Sudhakar Singh <sudhakars@nvidia.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Tim Moon <4406448+timmoon10@users.noreply.github.com>

fa91ed72

16 Jul, 2025 5 commits

Handle dtypes more carefully in multi-tensor Adam (#1888) · bda29934

Tim Moon authored Jul 16, 2025



* Add dtype checks in multi-tensor Adam
Signed-off-by: Tim Moon <tmoon@nvidia.com>

* Avoid throwing exception in destructor
Signed-off-by: Tim Moon <tmoon@nvidia.com>

* Debug test failures
Signed-off-by: Tim Moon <tmoon@nvidia.com>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci



---------
Signed-off-by: Tim Moon <tmoon@nvidia.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

bda29934

[Pytorch] Dynamo ONNX export support (#1497) · 0a1499fa

Paweł Gadziński authored Jul 16, 2025



* some initial code
Signed-off-by: Pawel Gadzinski <pgadzinski@nvidia.com>

* onnx support
Signed-off-by: Pawel Gadzinski <pgadzinski@nvidia.com>

* mxfp8 support
Signed-off-by: Pawel Gadzinski <pgadzinski@nvidia.com>

* fixed returning layernorm etc
Signed-off-by: Pawel Gadzinski <pgadzinski@nvidia.com>

* formatting
Signed-off-by: Pawel Gadzinski <pgadzinski@nvidia.com>

* lint fix
Signed-off-by: Pawel Gadzinski <pgadzinski@nvidia.com>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci



* license fix
Signed-off-by: Pawel Gadzinski <pgadzinski@nvidia.com>

* fix
Signed-off-by: Pawel Gadzinski <pgadzinski@nvidia.com>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci



* tests passing
Signed-off-by: Pawel Gadzinski <pgadzinski@nvidia.com>

* fix
Signed-off-by: Pawel Gadzinski <pgadzinski@nvidia.com>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci



* refactor
Signed-off-by: Pawel Gadzinski <pgadzinski@nvidia.com>

* lint
Signed-off-by: Pawel Gadzinski <pgadzinski@nvidia.com>

* fixes
Signed-off-by: Pawel Gadzinski <pgadzinski@nvidia.com>

* fix
Signed-off-by: Pawel Gadzinski <pgadzinski@nvidia.com>

* added pip install to test.sh
Signed-off-by: Pawel Gadzinski <pgadzinski@nvidia.com>

* Update transformer_engine/pytorch/export.py
Co-authored-by: Kirthi Shankar Sivamani <ksivamani@nvidia.com>
Signed-off-by: Paweł Gadziński <62263673+pggPL@users.noreply.github.com>

* fix
Signed-off-by: Pawel Gadzinski <pgadzinski@nvidia.com>

* fix
Signed-off-by: Pawel Gadzinski <pgadzinski@nvidia.com>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci



* float8currentscaling quantizer exception
Signed-off-by: Pawel Gadzinski <pgadzinski@nvidia.com>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci



* added to wheels
Signed-off-by: Pawel Gadzinski <pgadzinski@nvidia.com>

* onnx versions
Signed-off-by: Pawel Gadzinski <pgadzinski@nvidia.com>

* fix
Signed-off-by: Pawel Gadzinski <pgadzinski@nvidia.com>

* installations in tests
Signed-off-by: Pawel Gadzinski <pgadzinski@nvidia.com>

* fix
Signed-off-by: root <root@prenyx0221.a51.clusters.nvidia.com>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci



* lint fix
Signed-off-by: Pawel Gadzinski <pgadzinski@nvidia.com>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci



* fix
Signed-off-by: root <pgadzinski@nvidia.com>

* fixes
Signed-off-by: root <pgadzinski@nvidia.com>

* fixes
Signed-off-by: root <pgadzinski@nvidia.com>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci



* fixes
Signed-off-by: Pawel Gadzinski <pgadzinski@nvidia.com>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci



* fixes
Signed-off-by: Pawel Gadzinski <pgadzinski@nvidia.com>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci



* fix
Signed-off-by: Pawel Gadzinski <pgadzinski@nvidia.com>

* Update setup.py
Co-authored-by: Kirthi Shankar Sivamani <ksivamani@nvidia.com>
Signed-off-by: Paweł Gadziński <62263673+pggPL@users.noreply.github.com>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci



* onnxscript version chnage
Signed-off-by: Pawel Gadzinski <pgadzinski@nvidia.com>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci



* fixes
Signed-off-by: Pawel Gadzinski <pgadzinski@nvidia.com>

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci



* Fix CI
Signed-off-by: Kirthi Shankar Sivamani <ksivamani@nvidia.com>

* fix
Signed-off-by: Pawel Gadzinski <pgadzinski@gmail.com>

* Update build.yml
Signed-off-by: Kirthi Shankar Sivamani <ksivamani@nvidia.com>
Signed-off-by: Kirthi Shankar Sivamani <ksivamani@nvidia.com>

* Update pytorch.py
Signed-off-by: Kirthi Shankar Sivamani <ksivamani@nvidia.com>
Signed-off-by: Kirthi Shankar Sivamani <ksivamani@nvidia.com>

---------
Signed-off-by: Pawel Gadzinski <pgadzinski@nvidia.com>
Signed-off-by: Paweł Gadziński <62263673+pggPL@users.noreply.github.com>
Signed-off-by: root <root@prenyx0221.a51.clusters.nvidia.com>
Signed-off-by: root <pgadzinski@nvidia.com>
Signed-off-by: Kirthi Shankar Sivamani <ksivamani@nvidia.com>
Signed-off-by: Pawel Gadzinski <pgadzinski@gmail.com>
Co-authored-by: Kirthi Shankar Sivamani <ksivamani@nvidia.com>
Co-authored-by: root <root@prenyx0221.a51.clusters.nvidia.com>
Co-authored-by: Pawel Gadzinski <pgadzinski@gmail.com>

0a1499fa

[JAX] Support Flax sharding constraints (#1933) · c0c12e20

jberchtold-nvidia authored Jul 16, 2025



* Support flax sharding constraints
Signed-off-by: Jeremy Berchtold <jberchtold@nvidia.com>

* Add warning for deprecated TE logical axes
Signed-off-by: Jeremy Berchtold <jberchtold@nvidia.com>

* Update examples
Signed-off-by: Jeremy Berchtold <jberchtold@nvidia.com>

---------
Signed-off-by: Jeremy Berchtold <jberchtold@nvidia.com>

c0c12e20

[DCU] surpport NVTE_USE_HIPBLASLT_GROUPEDGEMM · 9406ff31
yuguo authored Jul 16, 2025

9406ff31
Bump up FA to 2.8.1 (#1949) · 6c526794
vcherepanov-nv authored Jul 15, 2025
```
Signed-off-by: Vladimir Cherepanov <vcherepanov@nvidia.com>
```
6c526794

15 Jul, 2025 2 commits
- Fix pytorch module import error · 148b5bea
  wenjh authored Jul 15, 2025
```
Signed-off-by: wenjh <wenjh@sugon.com>
```
  148b5bea
- [DCU] surpport channelwise int8 train · 793e0103
  yuguo authored Jul 15, 2025
  
  793e0103