Merge pull request #66 from facefusion/remove-redundant-encoder-calc

Remove redundant encoder calculation

face_swapper/src/exporting.py

@@ -1,7 +1,6 @@
 import os
 from configparser import ConfigParser
 
-
 import torch
 
 from .training import FaceSwapperTrainer

face_swapper/src/inferencing.py

@@ -4,7 +4,7 @@ import torch
 from torchvision import io
 
 from .helper import calc_embedding
-from .models.generator import Generator
+from .training import FaceSwapperTrainer
 
 CONFIG_PARSER = configparser.ConfigParser()
 CONFIG_PARSER.read('config.ini')
@@ -17,10 +17,7 @@ def infer() -> None:
 	config_target_path = CONFIG_PARSER.get('inferencing', 'target_path')
 	config_output_path = CONFIG_PARSER.get('inferencing', 'output_path')
 
-	state_dict = torch.load(config_generator_path).get('state_dict').get('generator')
-	generator = Generator(CONFIG_PARSER)
-	generator.load_state_dict(state_dict)
-	generator.eval()
+	generator = FaceSwapperTrainer.load_from_checkpoint(config_generator_path, config_parser = CONFIG_PARSER, map_location = 'cpu').eval()
 	embedder = torch.jit.load(config_embedder_path, map_location = 'cpu').eval()
 
 	source_tensor = io.read_image(config_source_path)

face_swapper/src/models/generator.py

@@ -19,8 +19,7 @@ class Generator(nn.Module):
 		self.generator.apply(init_weight)
 		self.masker.apply(init_weight)
 
-	def forward(self, source_embedding : Embedding, target_tensor : Tensor) -> Tuple[Tensor, Mask]:
-		target_features = self.encode_features(target_tensor)
+	def forward(self, source_embedding : Embedding, target_tensor : Tensor, target_features : Tuple[Feature, ...]) -> Tuple[Tensor, Mask]:
 		output_tensor = self.generator(source_embedding, target_features)
 		target_feature = target_features[-1]
 		output_mask = self.masker(target_tensor, target_feature)

face_swapper/src/training.py

@@ -53,7 +53,8 @@ class FaceSwapperTrainer(LightningModule):
 
 	def forward(self, source_embedding : Embedding, target_tensor : Tensor) -> Tuple[Tensor, Mask]:
 		with torch.no_grad():
-			output_tensor, output_mask = self.generator(source_embedding, target_tensor)
+			generator_target_features = self.generator.encode_features(target_tensor)
+			output_tensor, output_mask = self.generator(source_embedding, target_tensor, generator_target_features)
 
 		return output_tensor, output_mask
 
@@ -89,8 +90,8 @@ class FaceSwapperTrainer(LightningModule):
 		generator_optimizer, discriminator_optimizer = self.optimizers() #type:ignore[attr-defined]
 
 		source_embedding = calc_embedding(self.embedder, source_tensor, (0, 0, 0, 0))
-		generator_output_tensor, generator_output_mask = self.generator(source_embedding, target_tensor)
 		generator_target_features = self.generator.encode_features(target_tensor)
+		generator_output_tensor, generator_output_mask = self.generator(source_embedding, target_tensor, generator_target_features)
 		generator_output_features = self.generator.encode_features(generator_output_tensor)
 		discriminator_output_tensors = self.discriminator(generator_output_tensor)
 		adversarial_loss, weighted_adversarial_loss = self.adversarial_loss(discriminator_output_tensors)
@@ -138,7 +139,7 @@ class FaceSwapperTrainer(LightningModule):
 	def validation_step(self, batch : Batch, batch_index : int) -> Tensor:
 		source_tensor, target_tensor = batch
 		source_embedding = calc_embedding(self.embedder, source_tensor, (0, 0, 0, 0))
-		output_tensor, _ = self.generator(source_embedding, target_tensor)
+		output_tensor, _ = self.forward(source_embedding, target_tensor)
 		output_embedding = calc_embedding(self.embedder, output_tensor, (0, 0, 0, 0))
 		validation_score = (nn.functional.cosine_similarity(source_embedding, output_embedding).mean() + 1) * 0.5
 		self.log('validation_score', validation_score, sync_dist = True, prog_bar = True)