Updated optimization

README.md

@@ -24,14 +24,13 @@ The pretrained weights for most GANs are downloaded automatically. For those tha
 
 There are also some standard dependencies:
  - PyTorch (tested on version 1.7.1, but should work on any version)
- - Hydra 1.1
- - Albumentations, Retry
- - [Optional] Weights and Biases 
+ - [Hydra](https://github.com/facebookresearch/hydra) 1.1
+ - [Albumentations](https://github.com/albumentations-team/albumentations), [Kornia](https://github.com/kornia/kornia), [Retry](https://github.com/invl/retry)
+ - [Optional] [Weights and Biases](https://wandb.ai/)
 
  Install them with:
  ```bash
-pip install hydra-core --pre
-pip install albumentations tqdm retry tensorboard
+pip install hydra-core==1.1.0dev5 albumentations tqdm retry kornia
  ```
 
 
@@ -151,7 +150,7 @@ _Note:_ All commands are called from within the `src` directory.
 
 In the example commands below, we use BigBiGAN. You can easily switch out BigBiGAN for another model if you would like to. 
 
-**Optimizaion**
+**Optimization**
 ```bash
 PYTHONPATH=. python optimization/main.py data_gen/generator=bigbigan name=NAME
 ```

src/models/__init__.py

@@ -1,2 +1 @@
 from .unet_model import UNet
-from .ensemble import Ensemble

src/optimization/main.py

@@ -11,7 +11,7 @@
 from models.latent_shift_model import MODELS
 from optimization import utils
 
-# from pytorch_pretrained_gans import make_gan
+from pytorch_pretrained_gans import make_gan
 
 
 class UnsupervisedSegmentationLoss(torch.nn.Module):
@@ -65,7 +65,7 @@ def run(cfg: DictConfig):
     device = torch.device('cuda')
 
     # Load GAN
-    G = make_gan(gan_type=cfg.generator.gan_type, **cfg.generator.kwargs)
+    G = make_gan(gan_type=cfg.data_gen.generator.gan_type, **cfg.data_gen.generator.kwargs)
     G.eval().to(device)
     utils.set_requires_grad(G, False)
 
@@ -86,7 +86,7 @@ def run(cfg: DictConfig):
     scheduler = Scheduler(optimizer, **cfg.scheduler.kwargs)
 
     # Loss function
-    criterion = UnsupervisedSegmentationLoss(cfg.losses, image_size=cfg.generator.image_size)
+    criterion = UnsupervisedSegmentationLoss(cfg.losses, image_size=cfg.data_gen.generator.image_size)
     criterion.to(device)
 
     # Fixed vectors for visualization
@@ -138,8 +138,7 @@ def run(cfg: DictConfig):
         # Visualize with Tensorboard and save to file
         if i % cfg.vis_every == 0:
             img_grid = utils.create_grid(G=G, model=model, zs=z_vis_fixed, ys=y_vis_fixed, n_imgs=8)
-            img_file = "images" / f"{i}.png"
-            img_file.parent.mkdir(parents=True, exist_ok=True)
+            img_file = f"visualization-{i:05d}.png"
             img_grid = (img_grid * 255).astype(np.uint8)
             Image.fromarray(img_grid).save(str(img_file))
 

src/optimization/utils.py

@@ -1,3 +1,4 @@
+from pathlib import Path
 import random
 import numpy as np
 import torch
@@ -30,6 +31,7 @@ def create_grid(G, model, zs, ys, n_imgs=8, rs=[0, 2, 4, 8]):
 
 
 def save_checkpoint(model, checkpoint_dir='.', name='latest.pth', **kwargs):
+    checkpoint_dir = Path(checkpoint_dir)
     checkpoint_dir.mkdir(exist_ok=True)
     torch.save(dict(
         state_dict=model.state_dict(),