Torch models to pytorch models, bug fix, CPU support, etc #18
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…layers according to the origin paper.
xerebz
suggested changes
Feb 25, 2018
demo.py
Outdated
| parser.add_argument('--decoder3', default='./models/feature_invertor_conv3_1_mask.t7', help='Path to the decoder3') | ||
| parser.add_argument('--decoder2', default='./models/feature_invertor_conv2_1_mask.t7', help='Path to the decoder2') | ||
| parser.add_argument('--decoder1', default='./models/feature_invertor_conv1_1_mask.t7', help='Path to the decoder1') | ||
| parser.add_argument('--model', default='./PhotoWCTModels/photo_wct.pth', help='Path to the PhotoWCT model') |
There was a problem hiding this comment.
For ease of use, change this to something like: help='Path to the PhotoWCT model. These are provided by the PhotoWCT submodule, please use git submodule update --init to pull.'
demo.py
Outdated
| @@ -27,7 +20,8 @@ | |||
| args = parser.parse_args() | |||
|
|
|||
| # Load model | |||
| p_wct = PhotoWCT(args) | |||
| p_wct = PhotoWCT() | |||
| p_wct.load_state_dict(torch.load(args.model)) | |||
There was a problem hiding this comment.
For ease of use, use a try/except here with the exception message reminding the user to pull the submodule.
mingyuliutw
approved these changes
Feb 27, 2018
|
Hi, I think this PR is heavy and it's better to merge it before more PRs come in which may cause unexpected conflicts. @mingyuliutw @Yijunmaverick |
mingyuliutw
approved these changes
Mar 2, 2018
mingyuliutw
approved these changes
Mar 2, 2018
mingyuliutw
approved these changes
Mar 2, 2018
mingyuliutw
approved these changes
Mar 2, 2018
mingyuliutw
approved these changes
Mar 2, 2018
mingyuliutw
approved these changes
Mar 2, 2018
|
Merge #18 |
|
Processing speed and GPU performance memory wise is a bit worst than the original, was this supposed to improve GPU memory usage and overall time requirement too? |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR:
Convert torch models to pytorch models (listed in TODOs in the origin code) and
converter.pyshows how it was done. The pytorch model leaves in the submodulePhotoWCTModelswhich makes it easier to download from a server as Move trained models out of Google Drive #15 suggested.The models are refactored into less and clear classes. The layers are named according to the origin paper.
Fix a bug in
Propagator. It fails to process images with alpha channels because it does not open them with RGB mode.CPU support for PhotoWCT. PhotoWCT can work in CPU mode without using
.cuda(). This could make it ~10x slower (not too slow yet) but more friendly for those without GPUs or GPUs with less memory as Look at the memory consumption: should be able to process HD size picture #17I'm sorry that some codes in
photo_wct.pyare changed by the (PEP8) code formatter, so not too much of them are actually modified.Current code are tested. They can work well as before.