From ce1eed4f44c1a66c7729e993721c0c428e1f7332 Mon Sep 17 00:00:00 2001
From: andreasjansson <andreas@replicate.ai>
Date: Tue, 15 Mar 2022 17:38:42 -0700
Subject: [PATCH] Upgrade to Cog version 0.1

The [new version of Cog](https://github.com/replicate/cog/releases/tag/v0.1.0)
improves the Python API, along with several other changes.

This PR upgrades CLIPstyler to Cog version >= 0.1. I have already pushed
this to Replicate for you, so you don't need to do anything for the
demo to keep working: https://replicate.com/paper11667/clipstyler
---
 predict.py | 30 ++++++++++++++++--------------
 1 file changed, 16 insertions(+), 14 deletions(-)

diff --git a/predict.py b/predict.py
index f83d0e7..14328f6 100644
--- a/predict.py
+++ b/predict.py
@@ -1,5 +1,5 @@
 import sys
-from pathlib import Path
+from typing import Iterator
 import torch
 import torch.optim as optim
 from torchvision import transforms, models
@@ -11,11 +11,11 @@
 from template import imagenet_templates
 from torchvision.utils import save_image
 from torchvision.transforms.functional import adjust_contrast
-import cog
+from cog import BasePredictor, Input, Path
 from argparse import Namespace
 
 
-class Predictor(cog.Predictor):
+class Predictor(BasePredictor):
     def setup(self):
         self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
         self.VGG = models.vgg19(pretrained=True).features
@@ -26,10 +26,12 @@ def setup(self):
         self.style_net.to(self.device)
         self.clip_model, preprocess = clip.load('ViT-B/32', self.device, jit=False)
 
-    @cog.input("image", type=Path, help="Input image (will be cropped before style transfer)")
-    @cog.input("text", type=str, help="text for style transfer")
-    @cog.input("iterations", type=int, default=100, help="training iterations")
-    def predict(self, image, text, iterations):
+    def predict(
+        self,
+        image: Path = Input(description="Input image (will be cropped before style transfer)"),
+        text: str = Input(description="text for style transfer"),
+        iterations: int = Input(default=100, description="training iterations")
+    ) -> Iterator[Path]:
         training_args = {
             "lambda_tv": 2e-3,
             "lambda_patch": 9000,
@@ -141,17 +143,17 @@ def predict(self, image, text, iterations):
             if epoch % 20 == 0 or epoch == steps:
                 yield checkin(epoch, target, total_loss, content_loss, loss_patch, loss_glob, reg_tv, out_path)
 
-        return out_path
+        yield out_path
 
 
 @torch.no_grad()
 def checkin(epoch, target, total_loss, content_loss, loss_patch, loss_glob, reg_tv, out_path):
-    sys.stderr.write(f'After {epoch} iterations')
-    sys.stderr.write(f'Total loss: {total_loss.item()}')
-    sys.stderr.write(f'Content loss: {content_loss.item()}')
-    sys.stderr.write(f'patch loss: {loss_patch.item()}')
-    sys.stderr.write(f'dir loss: {loss_glob.item()}')
-    sys.stderr.write(f'TV loss: {reg_tv.item()}')
+    sys.stderr.write(f'After {epoch} iterations\n')
+    sys.stderr.write(f'Total loss: {total_loss.item()}\n')
+    sys.stderr.write(f'Content loss: {content_loss.item()}\n')
+    sys.stderr.write(f'patch loss: {loss_patch.item()}\n')
+    sys.stderr.write(f'dir loss: {loss_glob.item()}\n')
+    sys.stderr.write(f'TV loss: {reg_tv.item()}\n')
     output_image = target.clone()
     output_image = torch.clamp(output_image, 0, 1)
     output_image = adjust_contrast(output_image, 1.5)