Commits · 46e6342e84b9d44f2776c26bf38f1e124c5728df · OpenDAS / FastMoE

12 Jan, 2021 5 commits
- fix argument count bug in moe_test.py · 46e6342e
  Rick Ho authored Jan 12, 2021
  
  46e6342e
- add nccl flag in setup.py · 3501fa41
  Rick Ho authored Jan 12, 2021
  
  3501fa41
- Merge branch 'master' of github.com:xptree/poly-xfmr · dbf55073
  Jiezhong Qiu authored Jan 12, 2021
  
  dbf55073
- update · 80a862b4
  Jiezhong Qiu authored Jan 12, 2021
  
  80a862b4
- Merge pull request #1 from xptree/laekov/multigpu · b8aa893b
  Jiezhong Qiu authored Jan 12, 2021
```
Faster MoE implementation for both single GPU and multiple GPUs
```
  b8aa893b
11 Jan, 2021 5 commits
- multi-gpu performance test · b9c28810
  Rick Ho authored Jan 11, 2021
  
  b9c28810
- distributed test weight · 76327544
  Rick Ho authored Jan 11, 2021
  
  76327544
- multi-gpu forward pass test · 864a4522
  Rick Ho authored Jan 11, 2021
  
  864a4522
- make moe run with cuda · 069cf01a
  Rick Ho authored Jan 11, 2021
  
  069cf01a
- split function file · a4f7f1da
  Rick Ho authored Jan 11, 2021
  
  a4f7f1da
10 Jan, 2021 4 commits
- global scatter gather kernels and pytorch C function · ec322e4b
  Rick Ho authored Jan 10, 2021
  
  ec322e4b
- use expert_count array instead of expert_n · 7a2ad4a1
  Rick Ho authored Jan 10, 2021
  
  7a2ad4a1
- Merge branch 'laekov/batching' into laekov/multigpu · 307e0ad9
  Rick Ho authored Jan 10, 2021
  
  307e0ad9
- remove weight input to c expert count function · 861b75c1
  Rick Ho authored Jan 10, 2021
  
  861b75c1
09 Jan, 2021 11 commits
- stream manager across multiple devices · 60b93e39
  Rick Ho authored Jan 09, 2021
  
  60b93e39
- make backward pass test · 27d8beaa
  Rick Ho authored Jan 09, 2021
  
  27d8beaa
- moe backward (cannot pass test) · 92f1774a
  Rick Ho authored Jan 09, 2021
  
  92f1774a
- modify performance test · c91dfad8
  Rick Ho authored Jan 09, 2021
  
  c91dfad8
- split operators and make forward run · 49732231
  Rick Ho authored Jan 09, 2021
  
  49732231
- update stream manager · 143e21cc
  Rick Ho authored Jan 09, 2021
  
  143e21cc
- stream manager fixed · 2565f2fa
  Rick Ho authored Jan 09, 2021
  
  2565f2fa
- fix bugs to pass test · b8a212ef
  Rick Ho authored Jan 09, 2021
  
  b8a212ef
- Merge branch 'master' into laekov/batching · bf4388c0
  Rick Ho authored Jan 09, 2021
  
  bf4388c0
- degrade to single fc fwd · 284f1424
  Rick Ho authored Jan 09, 2021
  
  284f1424
- fix last indent bug · cbd86de8
  Rick Ho authored Jan 09, 2021
  
  cbd86de8
04 Jan, 2021 1 commit
- stream manager object instead of pointer · ef83c893
  Jiezhong Qiu authored Jan 04, 2021
  
  ef83c893
03 Jan, 2021 8 commits
- fix memory leak · e52c0380
  Jiezhong Qiu authored Jan 03, 2021
  
  e52c0380
- 2 out of 32 experts · 6429fa29
  Jiezhong Qiu authored Jan 03, 2021
  
  6429fa29
- cudafree malloced memory · 1bde1726
  Jiezhong Qiu authored Jan 03, 2021
  
  1bde1726
- can train now · 97a2b1be
  Jiezhong Qiu authored Jan 03, 2021
  
  97a2b1be
- update · 3be6ac10
  Jiezhong Qiu authored Jan 03, 2021
  
  3be6ac10
- update · 6212aa28
  Jiezhong Qiu authored Jan 03, 2021
  
  6212aa28
- make cudaStreamManager thread local · 27af1828
  Jiezhong Qiu authored Jan 03, 2021
  
  27af1828
- test data parallel (FAILED!) · f5cc759c
  Jiezhong Qiu authored Jan 03, 2021
  
  f5cc759c
02 Jan, 2021 6 commits
- fix bug · 7945bce6
  Jiezhong Qiu authored Jan 03, 2021
  
  7945bce6
- CUDA error at moe_cuda_kernel.cu:86 code=13(CUBLAS_STATUS_EXECUTION_FAILED) · 85306aa7
  Jiezhong Qiu authored Jan 03, 2021
  
  85306aa7
- update · 20cc924b
  Jiezhong Qiu authored Jan 03, 2021
  
  20cc924b
- set device according to input · 046455a8
  Jiezhong Qiu authored Jan 03, 2021
  
  046455a8
- incoporate customized cuda moe input xfmr-xl · 2338a26e
  Jiezhong Qiu authored Jan 03, 2021
  
  2338a26e
- add moe debug flag · d036918c
  Jiezhong Qiu authored Jan 02, 2021
  
  d036918c