Commits · 378ce1e144b0a09574b2f3e02ec4052ab2b68030 · OpenDAS / apex

11 Jun, 2018 4 commits
- Merge pull request #9 from NVIDIA/imagenet_fix · 378ce1e1
  mcarilli authored Jun 11, 2018
```
DDP fix, imagenet speedup
```
  378ce1e1
- [Imagenet example] Switch validation to same I/O pipeline so validation produces correct result. · 06ee98c2
  Christian Sarofeen authored Jun 08, 2018
  
  06ee98c2
- Switched to non-bucketed comm · 421c9e66
  Christian Sarofeen authored Jun 08, 2018
  
  421c9e66
- Better shortcut for shared_param = True · 137d822b
  Christian Sarofeen authored Jun 08, 2018
  
  137d822b
08 Jun, 2018 1 commit
- Compilation succeeds on 0.4, 18.04-6 containers, and current upstream master · fb075b86
  Michael Carilli authored Jun 08, 2018
  
  fb075b86
07 Jun, 2018 2 commits
- Reverting setup.py to rely on cpp_extension's native CUDA_HOME, because... · bf855389
  Michael Carilli authored Jun 07, 2018
```
Reverting setup.py to rely on cpp_extension's native CUDA_HOME, because upstream fixed cross-compilation
```
  bf855389
- Updating setup.py to enable cross-compilation · 70b33770
  Michael Carilli authored Jun 06, 2018
  
  70b33770
06 Jun, 2018 4 commits
- Merge changes from Christian's fork · af431d28
  Michael Carilli authored Jun 06, 2018
  
  af431d28
- Updating latest amp changes to use new C++ backend · d6db91a4
  Michael Carilli authored Jun 06, 2018
  
  d6db91a4
- Update README · 89564b69
  Michael Carilli authored Jun 06, 2018
  
  89564b69
- Macros based on torch.__version__ to compile with 0.4 and 0.5 · d506eff2
  Michael Carilli authored Jun 06, 2018
  
  d506eff2
05 Jun, 2018 2 commits
- Merging latest master · 61b452e8
  Michael Carilli authored Jun 05, 2018
  
  61b452e8
- remove cuda check in build_cffi to handle docker installs · db6ae13a
  Carl Case authored Jun 05, 2018
  
  db6ae13a
01 Jun, 2018 3 commits
- Merge pull request #7 from pooyadavoodi/world_size_multiproc · 6143b30f
  Christian Sarofeen authored Jun 01, 2018
```
Read world_size argument in multiproc if specified in argslist
```
  6143b30f
- Merge pull request #8 from pooyadavoodi/speed_metric · eb0e59eb
  Christian Sarofeen authored Jun 01, 2018
```
Print speed (img/sec) in imagenet example
```
  eb0e59eb
- Print speed (img/sec) in imagenet example · dcda3b56
  Pooya Davoodi authored Jun 01, 2018
  
  dcda3b56
31 May, 2018 1 commit
- Read world_size argument in multiproc if specified in argslist · 38ea0e90
  Pooya Davoodi authored May 31, 2018
  
  38ea0e90
30 May, 2018 7 commits
- readme · 5ae9fbc4
  Carl Case authored May 30, 2018
  
  5ae9fbc4
- Hard ban on fp16 BCELoss · fb419005
  Carl Case authored May 30, 2018
  
  fb419005
- rename build->init · 6d30e1ff
  Carl Case authored May 30, 2018
  
  6d30e1ff
- WIP: better annotation / user function registry support · ec431d33
  Carl Case authored May 30, 2018
  
  ec431d33
- support multi-loss scaling per-optimizer correctly · 614b11ff
  Carl Case authored May 29, 2018
  
  614b11ff
- bugfix: keep cache up-to-date on parameter require_grad-ness · 8be1b053
  Carl Case authored May 29, 2018
  
  8be1b053
- Optimizer wrapper; loss scaling class; no-op handle; start multi-loss · 9cc74429
  Carl Case authored May 24, 2018
  
  9cc74429
27 May, 2018 1 commit
- Re-add normalization, correct typing. · 414dc119
  Christian Sarofeen authored May 27, 2018
  
  414dc119
26 May, 2018 1 commit
- Fleshed out Cuda version checking and compiling for multiple arches · fb7d4e1d
  Michael Carilli authored May 25, 2018
  
  fb7d4e1d
25 May, 2018 3 commits
- Decrease default message size. · 0081afb8
  Christian Sarofeen authored May 25, 2018
  
  0081afb8
- Update imagenet example to fast version. · cae6005c
  Christian Sarofeen authored May 25, 2018
  
  cae6005c
- Transferred backend and build system to use Pytorch C++ extension + ATen dispatch. · d17a015f
  Michael Carilli authored May 25, 2018
  
  d17a015f
24 May, 2018 1 commit
- 1804 RNN fix · ea93767d
  Carl Case authored May 24, 2018
  
  ea93767d
23 May, 2018 5 commits
- Revert "Revert distributed to simpler version until backward hooks are fixed." · 343590a1
  Christian Sarofeen authored May 23, 2018
```
This reverts commit 47ac5c2b.
```
  343590a1
- Merge pull request #4 from NVIDIA/amp_compat_fix · 1737ce1e
  mcarilli authored May 23, 2018
```
Fix compatibility checks for 18.04 container
```
  1737ce1e
- fix compatibility checks for 18.04 container · 9ce3a33d
  Carl Case authored May 23, 2018
  
  9ce3a33d
- Actually fix builds for pre 0.4 · ee117aa8
  Christian Sarofeen authored May 23, 2018
  
  ee117aa8
- Fix pre 0.4 builds. · 4f8829f9
  Christian Sarofeen authored May 23, 2018
```
Fix regex expression to search for libaten
```
  4f8829f9
22 May, 2018 3 commits
- Merge pull request #3 from raulpuric/patch-2 · 70d2e3b7
  Christian Sarofeen authored May 22, 2018
```
Create LARC.py
```
  70d2e3b7
- Create LARC.py · 889d871b
  Raul Puri authored May 22, 2018
  
  889d871b
- Add device sync in imagenet example. · 61c1e160
  Christian Sarofeen authored May 21, 2018
  
  61c1e160
21 May, 2018 1 commit
- Add device sync in imagenet example. · 2d5b71bd
  Christian Sarofeen authored May 21, 2018
  
  2d5b71bd
18 May, 2018 1 commit
- Initial support for automatic mixed precision · e733e78c
  Carl Case authored May 16, 2018
  
  e733e78c