Commits · 8113d9e0cdcf388544a425b6bba2d10f95400d89 · OpenDAS / TransformerEngine

18 Jul, 2025 6 commits
- Merge branch 'develop_v2.4' of http://10.16.6.30/dcutoolkit/deeplearing/TransformerEngine · 8113d9e0
  yuguo authored Jul 18, 2025
  
  8113d9e0
- [DCU] fix · d9847b6d
  yuguo authored Jul 18, 2025
  
  d9847b6d
- Merge branch 'develop_v2.4' of http://10.16.6.30/dcutoolkit/deeplearing/TransformerEngine · 93ecbc82
  yuguo authored Jul 18, 2025
  
  93ecbc82
- [DCU] fix bias gradient · 736e8f8b
  yuguo authored Jul 18, 2025
  
  736e8f8b
- Merge branch 'develop_v2.4' · bb8cf71b
  wenjh authored Jul 18, 2025
  
  bb8cf71b
- Optional lightop · f5349823
  wenjh authored Jul 18, 2025
```
Signed-off-by: wenjh <wenjh@sugon.com>
```
  f5349823
17 Jul, 2025 4 commits
- Merge branch 'develop_v2.4' · 429226fb
  wenjh authored Jul 17, 2025
  
  429226fb
- Use lightop replace w8a8_mutmal_extension · 1036ccfe
  wenjh authored Jul 17, 2025
```
Signed-off-by: wenjh <wenjh@sugon.com>
```
  1036ccfe
- Merge branch 'develop_v2.4' of http://10.16.6.30/dcutoolkit/deeplearing/TransformerEngine · 3b0a1009
  yuguo authored Jul 17, 2025
  
  3b0a1009
- [DCU] fix · 00738a42
  yuguo authored Jul 17, 2025
  
  00738a42
16 Jul, 2025 4 commits
- Merge branch 'develop_v2.4' of http://10.16.6.30/dcutoolkit/deeplearing/TransformerEngine · 686af9c3
  yuguo authored Jul 16, 2025
  
  686af9c3
- [DCU] surpport NVTE_USE_HIPBLASLT_GROUPEDGEMM · 9406ff31
  yuguo authored Jul 16, 2025
  
  9406ff31
- Merge branch 'develop_v2.4' · ee787b22
  wenjh authored Jul 16, 2025
  
  ee787b22
- Fix import error in test_batched_linear · bc2d9697
  wenjh authored Jul 16, 2025
```
Signed-off-by: wenjh <wenjh@sugon.com>
```
  bc2d9697
15 Jul, 2025 6 commits
- Merge branch 'develop_v2.4' · b32741e2
  wenjh authored Jul 15, 2025
  
  b32741e2
- Fix pytorch module import error · 148b5bea
  wenjh authored Jul 15, 2025
```
Signed-off-by: wenjh <wenjh@sugon.com>
```
  148b5bea
- Merge branch 'develop_v2.4' of http://10.16.6.30/dcutoolkit/deeplearing/TransformerEngine · a5892578
  yuguo authored Jul 15, 2025
  
  a5892578
- [DCU] surpport channelwise int8 train · 793e0103
  yuguo authored Jul 15, 2025
  
  793e0103
- Merge branch 'develop_v2.4' · f9faa7ca
  wenjh authored Jul 15, 2025
  
  f9faa7ca
- Fix import pytorch error · 3939e719
  wenjh authored Jul 15, 2025
```
Signed-off-by: wenjh <wenjh@sugon.com>
```
  3939e719
11 Jul, 2025 3 commits
- Merge branch 'develop_v2.4' · fdf60506
  wenjh authored Jul 11, 2025
  
  fdf60506
- Merge branch 'develop_v2.4' into w8a8_dev_v2.4 · 3b1f30a9
  wenjh authored Jul 11, 2025
  
  3b1f30a9
- Support w8a8_matmul_extension · 6a20ff90
  wenjh authored Jul 11, 2025
```
Signed-off-by: wenjh <wenjh@sugon.com>
```
  6a20ff90
09 Jul, 2025 4 commits
- Merge branch 'develop_v2.4' of http://10.16.6.30/dcutoolkit/deeplearing/TransformerEngine · 403db136
  yuguo authored Jul 09, 2025
  
  403db136
- [DCU] channelwise batchgemm for MOE · 76023d21
  yuguo authored Jul 09, 2025
  
  76023d21
- Merge branch 'develop_v2.4' · c3a36d7e
  wenjh authored Jul 09, 2025
  
  c3a36d7e
- Fix int8 gemm nt and wgrad · 5fcf30ba
  wenjh authored Jul 09, 2025
```
Signed-off-by: wenjh <wenjh@sugon.com>
```
  5fcf30ba
08 Jul, 2025 2 commits
- Merge branch 'develop_v2.4' of http://10.16.6.30/dcutoolkit/deeplearing/TransformerEngine · 800f4dff
  yuguo authored Jul 08, 2025
  
  800f4dff
- [DCU] Preliminary support for channelwise · 9fe13a33
  yuguo authored Jul 08, 2025
  
  9fe13a33
03 Jul, 2025 2 commits
- Merge branch 'develop_v2.4' · 1e018a45
  wenjh authored Jul 03, 2025
  
  1e018a45
- Fix kernel crash on block_len=64 · 40a4d896
  wenjh authored Jul 03, 2025
```
Signed-off-by: wenjh <wenjh@sugon.com>
```
  40a4d896
02 Jul, 2025 1 commit
- Resolve merge issues from develop_v2.4 · 4ef4eae6
  wenjh authored Jul 02, 2025
```
Signed-off-by: wenjh <wenjh@sugon.com>
```
  4ef4eae6
01 Jul, 2025 2 commits

Merge develop_v2.4 · 0e886dab
wenjh authored Jul 01, 2025
```
Signed-off-by: wenjh <wenjh@sugon.com>
```
0e886dab

[Blockwise] Add support block_len=64 support · b944277c

wenjh authored Jun 25, 2025



Add env to chose blocklen of blockwise quantize.
Signed-off-by: wenjh <wenjh@sugon.com>

Fix pytest of blockwise error
Signed-off-by: wenjh <wenjh@sugon.com>

Resolve new api in  int8 gemm test
Signed-off-by: wenjh <wenjh@sugon.com>

Fix incorrect launch parm
Signed-off-by: wenjh <wenjh@sugon.com>

Fix 1D blockwise(64) acc error
Signed-off-by: wenjh <wenjh@sugon.com>

b944277c

20 Jun, 2025 5 commits
- Merge branch 'develop_v2.4' into 'main' · e56de127
  yuguo authored Jun 20, 2025
```
[DCU] fix megatron MOE int8 train bugs

See merge request dcutoolkit/deeplearing/TransformerEngine!37
```
  e56de127
- [DCU] fix megatron MOE int8 train bugs · 251dcc7e
  yuguo authored Jun 20, 2025
  
  251dcc7e
- [DCU] fix 2.5 compile issues · 68487b2a
  yuguo authored Jun 20, 2025
  
  68487b2a
- Merge branch 'develop_v2.4' into 'main' · 25e709aa
  yuguo authored Jun 20, 2025
```
[DCU] fix megatron MOE int train issues

See merge request dcutoolkit/deeplearing/TransformerEngine!36
```
  25e709aa
- [DCU] fix megatron MOE int train issues · 7640a8d4
  yuguo authored Jun 20, 2025
  
  7640a8d4
19 Jun, 2025 1 commit
- [DCU] fix TORCH_COMM_CU_NUMS conflicts in 2.5 · ffbef335
  yuguo authored Jun 19, 2025
  
  ffbef335