Commits · f103e2a3e0140a3bb2402b642099a124c12ee81b · OpenDAS / TransformerEngine

23 Jan, 2026 2 commits

[DCU] Remove redundant shared memory in rowwise kernel · f103e2a3

zc20020701 authored Jan 23, 2026


Signed-off-by: zhaochao <zhaochao1@sugon.com>

See merge request dcutoolkit/deeplearing/TransformerEngine!72
Co-authored-by: zhaochao <zhaochao1@sugon.com>

f103e2a3

Refine the constraints while using lightop in gemm.py · 0d2c4ca9
wenjh authored Jan 23, 2026
```
Signed-off-by: wenjh <wenjh@sugon.com>
```
0d2c4ca9

21 Jan, 2026 1 commit

Add NVTE_USE_LIGHTOP env var to control lightop import · 9ccc82eb

maxiao3 authored Jan 21, 2026


Signed-off-by: maxiao3 <maxiao3@sugon.com>

See merge request dcutoolkit/deeplearing/TransformerEngine!71

9ccc82eb

12 Jan, 2026 1 commit
- Fix building on nmz · 77e5b163
  wenjh authored Jan 12, 2026
```
Signed-off-by: wenjh <wenjh@sugon.com>
```
  77e5b163
09 Jan, 2026 2 commits
- Fix swizzle, swap_first_dims and RMSNorm issues · d80e8084
  wenjh authored Jan 09, 2026
```
Signed-off-by: wenjh <wenjh@sugon.com>
```
  d80e8084
- Fix tests of L0 test_numeric and L1 test_fusible_ops · 6377424c
  wenjh authored Jan 09, 2026
```
Signed-off-by: wenjh <wenjh@sugon.com>
```
  6377424c
07 Jan, 2026 1 commit
- Add nmz support · 847f4654
  wenjh authored Jan 07, 2026
```
Signed-off-by: wenjh <wenjh@sugon.com>
```
  847f4654
19 Dec, 2025 1 commit
- Add bias fwd/bwd at group gemm · 0b03f913
  wenjh authored Dec 19, 2025
```
Signed-off-by: wenjh <wenjh@sugon.com>
```
  0b03f913
15 Dec, 2025 1 commit
- Fix blaslt group gemm crush · 500272e8
  wenjh authored Dec 15, 2025
  
  500272e8
26 Nov, 2025 2 commits
- Merge branch 'develop_v2.8' into release_v2.8 · 1edc9e13
  wenjh authored Nov 26, 2025
  
  1edc9e13
- Merge branch 'fix_develop2.8_zc' into 'develop_v2.8' · 3a040217
  wenjh authored Nov 26, 2025
```
[DCU]Fix some bugs

See merge request dcutoolkit/deeplearing/TransformerEngine!56
```
  3a040217
12 Nov, 2025 2 commits
- Fix build error · a622988a
  wenjh authored Nov 12, 2025
  
  a622988a
- Fix hipblaslt handle manage · f4bd89eb
  wenjh authored Nov 12, 2025
  
  f4bd89eb
08 Nov, 2025 1 commit
- Fix user args core dump in mt · a13c52ad
  wenjh authored Nov 08, 2025
  
  a13c52ad
03 Nov, 2025 8 commits
- [DCU] fix some bugs in test_numerics.py · f7c66e28
  zhaochao authored Nov 03, 2025
  
  f7c66e28
- [DCU]Skip configurations that FlashAttention does not support · 87682fe2
  zhaochao authored Nov 03, 2025
```
Signed-off-by: zhaochao <zhaochao1@sugon.com>
```
  87682fe2
- [DCU]Resolve the issue of checkpoint test weights not existing. · 9d34e27a
  zhaochao authored Nov 03, 2025
```
Signed-off-by: zhaochao <zhaochao1@sugon.com>
```
  9d34e27a
- [DCU] Fix the bug in test_onnx_export.py under L0 · d5cd815f
  zhaochao authored Nov 03, 2025
```
Signed-off-by: zhaochao <zhaochao1@sugon.com>
```
  d5cd815f
- [DCU] Skip alpha non-1 tests · ef65dd33
  zhaochao authored Nov 03, 2025
```
Signed-off-by: zhaochao <zhaochao1@sugon.com>
```
  ef65dd33
- [DCU] fix bug with cannot import name 'use_lightop_w8a8' from 'transformer_engine.pytorch.utils' · 3d36696b
  zhaochao authored Nov 03, 2025
```
Signed-off-by: zhaochao <zhaochao1@sugon.com>
```
  3d36696b
- [DCU] Skip some tests in test_cuda_graphs.py under L0 · 2fc4b10c
  zhaochao authored Nov 03, 2025
```
Signed-off-by: zhaochao <zhaochao1@sugon.com>
```
  2fc4b10c
- [DCU] Skip some tests in test_sanity.py · 6af7b77d
  zhaochao authored Nov 03, 2025
```
Signed-off-by: zhaochao <zhaochao1@sugon.com>
```
  6af7b77d
31 Oct, 2025 1 commit

Merge branch 'TE_develop2.8' into 'develop_v2.8' · 3a5755b1

wenjh authored Oct 31, 2025

[DCU]Fix memory overflow and test-didistributed in L1_pytorch_istributed_unittest

See merge request dcutoolkit/deeplearing/TransformerEngine!49

3a5755b1

17 Oct, 2025 4 commits
- [DCU]Fix the original code · b11d6fca
  tabuchixiangcai3 authored Oct 17, 2025
```
Signed-off-by: Tangao <2205747538@qq.com>
```
  b11d6fca
- Merge branch 'develop_v2.8' of... · 5e7dd67e
  yuguo authored Oct 17, 2025
```
Merge branch 'develop_v2.8' of http://10.16.6.30/dcutoolkit/deeplearing/TransformerEngine into release_v2.8
```
  5e7dd67e
- Merge branch 'develop_v2.8' into 'develop_v2.8' · 4b65dfa3
  yuguo authored Oct 17, 2025
```
Update activation offload code to align with the official version

See merge request dcutoolkit/deeplearing/TransformerEngine!52
```
  4b65dfa3
- Update activation offload code to align with the official version · 9711d439
  dongcl authored Oct 17, 2025
  
  9711d439
16 Oct, 2025 4 commits
- Merge branch 'develop_v2.8' of... · 9df7fe23
  yuguo authored Oct 16, 2025
```
Merge branch 'develop_v2.8' of http://10.16.6.30/dcutoolkit/deeplearing/TransformerEngine into release_v2.8
```
  9df7fe23
- Merge branch 'develop_v2.8' of... · 712d526a
  yuguo authored Oct 16, 2025
```
Merge branch 'develop_v2.8' of http://10.16.6.30/dcutoolkit/deeplearing/TransformerEngine into develop_v2.8
```
  712d526a
- [DCU] remove redundant gemm · 47077129
  yuguo authored Oct 16, 2025
  
  47077129
- [DCU]Fix memory overflow and test-didistributed in L1_pytorch_istributed_unittest · 2a64c9a6
  tabuchixiangcai3 authored Oct 16, 2025
```
Signed-off-by: Tangao <2205747538@qq.com>
```
  2a64c9a6
15 Oct, 2025 8 commits
- Merge branch 'release_v2.8' of... · d7c42515
  yuguo authored Oct 15, 2025
```
Merge branch 'release_v2.8' of http://10.16.6.30/dcutoolkit/deeplearing/TransformerEngine into release_v2.8
```
  d7c42515
- [DCU] fix release 2.8 compile issues · 962b976d
  yuguo authored Oct 15, 2025
  
  962b976d
- Merge branch 'develop_v2.8' into release_v2.8 · 07f59c3d
  wenjh authored Oct 15, 2025
  
  07f59c3d
- Fix typo · a26a0c30
  wenjh authored Oct 15, 2025
```
Signed-off-by: wenjh <wenjh@sugon.com>
```
  a26a0c30
- Merge branch 'develop_v2.8' of... · 688c7ab9
  yuguo authored Oct 15, 2025
```
Merge branch 'develop_v2.8' of http://10.16.6.30/dcutoolkit/deeplearing/TransformerEngine into release_v2.8
```
  688c7ab9
- [DCU] fix compile issues · aa62d24c
  yuguo authored Oct 15, 2025
  
  aa62d24c
- Merge branch 'develop_v2.8' of... · d685b559
  yuguo authored Oct 15, 2025
```
Merge branch 'develop_v2.8' of http://10.16.6.30/dcutoolkit/deeplearing/TransformerEngine into release_v2.8
```
  d685b559
- [DCU] fix compile issues · 8d5cd8c6
  yuguo authored Oct 15, 2025
  
  8d5cd8c6
13 Oct, 2025 1 commit
- Attmp fix build error · cc554d59
  wenjh authored Oct 13, 2025
```
Signed-off-by: wenjh <wenjh@sugon.com>
```
  cc554d59