Commit Graph

1842 Commits

Author SHA1 Message Date
torzdf
c3c3483eed AMD Bugfixes: Save Optimizer Weights 2022-06-23 15:55:24 +01:00
torzdf
06468c97d4 Training: Add setting option to save optimizer weights 2022-06-23 14:54:09 +01:00
torzdf
66c2b7b979 Bugfix: Stats elapsed time 2022-06-22 19:20:51 +01:00
torzdf
4d42591ffc Mixed Precision bugfix for AMD 2022-06-22 12:49:07 +01:00
torzdf
0d2eafa485 Enable change of precision for existing models 2022-06-22 01:58:59 +01:00
torzdf
da942e3b3f Nan bugfix in Stats 2022-06-21 12:32:09 +01:00
torzdf
5d700e869e stats bugfixes 2022-06-20 13:07:37 +01:00
torzdf
ad408f07fb typofix 2022-06-20 12:46:23 +01:00
torzdf
9e94273f89 bigfix - Live graph reading 2022-06-20 12:04:35 +01:00
torzdf
e2fc070370 Bug fixes
- PhazeA tooltip spacing
  - Graph live cache bug
2022-06-20 03:38:47 +01:00
torzdf
d7ffcda3b0 Add LPIPS loss function 2022-06-19 18:42:34 +01:00
torzdf
bad5025aea Core updates
- Change loss loading mechanism
   - Autosize tooltips based on content size
   - Random linting + code modernisation
2022-06-19 12:32:39 +01:00
torzdf
f2e6f24651 Centralize model storage 2022-06-18 19:54:02 +01:00
torzdf
ef79a3d8cb Add AlexNet + SqueezeNet definitions 2022-06-18 18:21:41 +01:00
torzdf
1d434b73a4 Enable LogCosh Loss for AMD 2022-06-18 16:12:39 +01:00
torzdf
84b47fde95 linting 2022-06-18 15:36:39 +01:00
torzdf
76cb535e82 linting 2022-06-18 15:32:02 +01:00
torzdf
42407e9d5c linting 2022-06-18 14:52:18 +01:00
torzdf
308c5edf39 Loss updates
- Split Loss to own package
  - Add Focal Frequency Loss (Nvidia only)
  - Enable GMSD for AMD
2022-06-18 14:33:45 +01:00
torzdf
9d55ade27f amd gpu-stats bugfix 2022-06-18 11:16:03 +01:00
torzdf
d9c84a5f9f Add Laplacian Pyramid Loss 2022-06-18 02:29:19 +01:00
torzdf
04337e0c5e SSIM Updates
- Standardize DSSIM Function
  - Implement MSSIM function for AMD
2022-06-18 00:02:28 +01:00
torzdf
d02d51f2bb doc fixes 2022-06-17 22:01:57 +01:00
torzdf
94c3dcff7e Training updates
- Add multiple selected loss functions
  - Unlock loss as a model configuration
  - Phaze-A remove encoder scaling max xap
2022-06-17 20:24:26 +01:00
torzdf
98a65277d8 Fix AMD Tests + docs 2022-06-17 17:59:55 +01:00
torzdf
ff6b0209dd
Refactoring and TravisCI to Github Actions (#1239)
* refactor training

* travis to actions
2022-06-17 03:24:16 +01:00
torzdf
fbf8bdd72b Merge branch 'master' into apple_setup 2022-06-16 16:41:15 +01:00
torzdf
7e4bcbe36a typofix 2022-06-16 16:40:56 +01:00
torzdf
5e489f3b84 typing bugfix 2022-06-16 16:40:56 +01:00
torzdf
9818f6ee70 bugfix 2022-06-16 16:40:56 +01:00
torzdf
b039cbae69 bugfix 2022-06-16 16:40:56 +01:00
torzdf
cb7cb14e25 bugfix: Mask blend. Apply masks in sequence 2022-06-16 16:40:56 +01:00
torzdf
4f49de52e7 pin protobuf
Setup: automatically detect for apple-silicon
2022-06-16 16:40:49 +01:00
geewiz94
0f67790071 Some Apple Silicion setup fixes (#1238)
* Move pynvx to Nvidia requirements

* Add missing tensorflow-deps and fix tensorflow-metal version

* Fix setup not using Conda

* Update INSTALL.md
2022-06-16 16:40:49 +01:00
torzdf
a586ef6bf3 Add Apple M1 to setup.py
add libblas to requirements
2022-06-16 16:40:45 +01:00
torzdf
9563283502 typing bugfix 2022-06-16 10:02:12 +01:00
torzdf
493030f4b0 bugfix 2022-06-16 02:50:07 +01:00
torzdf
5559c54b76 bugfix 2022-06-16 02:27:07 +01:00
torzdf
db778bae89 bugfix: Mask blend. Apply masks in sequence 2022-06-16 02:19:38 +01:00
Rushi Chaudhari
199911ed2e upgrade Docker Ubuntu to 18.04
Update INSTALL.md
2022-06-13 01:00:52 +01:00
torzdf
917acaa452 bugfixes
- Fix MS-SSIM on multi-gpu
  - Swallow print bug on multi-gpu
2022-06-12 02:04:11 +01:00
torzdf
d933b6dee3 bugfix: Literal import for python<3.7 2022-06-09 20:16:20 +01:00
torzdf
828c96611b convert: fix mask casting bug 2022-06-08 21:12:17 +01:00
torzdf
3d8e674adc convert - Fix affine borders 2022-06-08 12:44:05 +01:00
torzdf
2298d51237 Preview Tool fixes
- Prevent errors from getting swallowed
  - Correctly handle None mask type
  - Consistent box blending
2022-06-08 12:08:18 +01:00
torzdf
73442b520c convert: remove box blend plugin 2022-06-08 00:54:47 +01:00
torzdf
df092470ec Convert - Add ability to directionally erode mask 2022-06-07 19:24:15 +01:00
torzdf
20a657d6cf Bump minimum TF Version 2022-06-07 10:54:51 +01:00
torzdf
5c9fa1aa03 Train updates
- Output full model summary last
  - Fix activations in Phaze-A upscales
  - Phaze-A Add option to place some upscales in fc model
2022-06-07 01:11:26 +01:00
torzdf
c7e577e887 Phaze-A - Updates for future implementations 2022-06-06 13:15:21 +01:00