Migrate to Pytorch 1.0 #16
Open
Conversation
This file contains hidden or 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
use torch.no_grad() instead of volatile variables. update example images to also use relu5_1 stylization. Point models download to new file. Remove model conversion script. add .gitignore
black-puppydog
force-pushed
the
master
branch
from
140d2e3 to
bb093b9
Compare
Author
|
hmmmm... I pushed some other changes that shouldn't go into this repo, or at least into a different file... |
Author
|
@eridgd yeah @sunshineatnoon did a great job with this one. |
The pascal server can't really be relied on long-term. This is not great form, but the models are a mere 120MB so this is workable, and since the repo is pretty much dormant I don't see a reason to go through more trouble.
since they're now included in the repo, no need to confuse users with "you need the models"
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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Outdated suggestions cannot be applied.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 takes the code all the way to pytorch 1.0:
I had to migrate the encoder/decoder models since
load_luaseems to have been removed from the API (?!) and while I was at it I also combined all the encoders into one, which saves a little bit of RAM and makes the file much more usable. So all those 6 models are now standard.pth.tarfiles to be used withnet.load_state_dict(torch.load(filename)).I changed all the models with a custom script (see intermediate commit) that can only be run with pytorch 0.4.1 then switched everything to those models and deleted the conversion script and the old model definitions.
On top of that I had to make some small changes to make the transition to the unified Tensor/Variable classes.
I also updated
Readme.mdto point at a zip file I prepared with the new models, of course that can be changed if you want to put this on google drive again.And I updated the example images, since you didn't do that when you included layer 5 into the model. I know you prefer the output without relu5_1, but I think keeping the visuals and the code consistent makes sense, no?