From 9f9f9dbad7086c888a34b522747c23d6b42e7a56 Mon Sep 17 00:00:00 2001 From: henryruhs Date: Wed, 25 Jun 2025 11:36:31 +0200 Subject: [PATCH] Rename calcXXX to calculateXXX --- hyperswap/src/helper.py | 2 +- hyperswap/src/inferencing.py | 4 ++-- hyperswap/src/models/loss.py | 14 +++++++------- hyperswap/src/training.py | 10 +++++----- 4 files changed, 15 insertions(+), 15 deletions(-) diff --git a/hyperswap/src/helper.py b/hyperswap/src/helper.py index 71ab19b..2abc70a 100644 --- a/hyperswap/src/helper.py +++ b/hyperswap/src/helper.py @@ -34,7 +34,7 @@ def convert_tensor(input_tensor : Tensor, convert_template : ConvertTemplate) -> return output_tensor -def calc_embedding(embedder : EmbedderModule, input_tensor : Tensor, padding : Padding) -> Embedding: +def calculate_embedding(embedder : EmbedderModule, input_tensor : Tensor, padding : Padding) -> Embedding: crop_tensor = convert_tensor(input_tensor, 'arcface_128_to_arcface_112_v2') crop_tensor = nn.functional.interpolate(crop_tensor, size = 112, mode = 'area') crop_tensor[:, :, :padding[0], :] = 0 diff --git a/hyperswap/src/inferencing.py b/hyperswap/src/inferencing.py index b9f0c54..3271ccf 100644 --- a/hyperswap/src/inferencing.py +++ b/hyperswap/src/inferencing.py @@ -3,7 +3,7 @@ import configparser import torch from torchvision import io -from .helper import calc_embedding +from .helper import calculate_embedding from .training import HyperSwapTrainer CONFIG_PARSER = configparser.ConfigParser() @@ -22,6 +22,6 @@ def infer() -> None: source_tensor = io.read_image(config_source_path) target_tensor = io.read_image(config_target_path) - source_embedding = calc_embedding(embedder, source_tensor, (0, 0, 0, 0)) + source_embedding = calculate_embedding(embedder, source_tensor, (0, 0, 0, 0)) output_tensor, _ = generator(source_embedding, target_tensor) io.write_jpeg(output_tensor, config_output_path) diff --git a/hyperswap/src/models/loss.py b/hyperswap/src/models/loss.py index 494e7f9..45a32a2 100644 --- a/hyperswap/src/models/loss.py +++ b/hyperswap/src/models/loss.py @@ -6,7 +6,7 @@ from pytorch_msssim import ssim from torch import Tensor, nn from torchvision import transforms -from ..helper import calc_embedding, dilate_mask +from ..helper import calculate_embedding, dilate_mask from ..types import EmbedderModule, FaceMaskerModule, Feature, GazerModule, Loss, Mask @@ -97,8 +97,8 @@ class ReconstructionLoss(nn.Module): def forward(self, source_tensor : Tensor, target_tensor : Tensor, output_tensor : Tensor) -> Tuple[Loss, Loss]: with torch.no_grad(): - source_embedding = calc_embedding(self.embedder, source_tensor, (0, 0, 0, 0)) - target_embedding = calc_embedding(self.embedder, target_tensor, (0, 0, 0, 0)) + source_embedding = calculate_embedding(self.embedder, source_tensor, (0, 0, 0, 0)) + target_embedding = calculate_embedding(self.embedder, target_tensor, (0, 0, 0, 0)) has_similar_identity = torch.cosine_similarity(source_embedding, target_embedding) > 0.8 @@ -120,8 +120,8 @@ class IdentityLoss(nn.Module): self.embedder = embedder def forward(self, source_tensor : Tensor, output_tensor : Tensor) -> Tuple[Loss, Loss]: - output_embedding = calc_embedding(self.embedder, output_tensor, (30, 0, 10, 10)) - source_embedding = calc_embedding(self.embedder, source_tensor, (30, 0, 10, 10)) + output_embedding = calculate_embedding(self.embedder, output_tensor, (30, 0, 10, 10)) + source_embedding = calculate_embedding(self.embedder, source_tensor, (30, 0, 10, 10)) identity_loss = (1 - torch.cosine_similarity(source_embedding, output_embedding)).mean() weighted_identity_loss = identity_loss * self.config_identity_weight return identity_loss, weighted_identity_loss @@ -169,7 +169,7 @@ class MaskLoss(nn.Module): self.mse_loss = nn.MSELoss() def forward(self, target_tensor : Tensor, output_mask : Mask) -> Tuple[Loss, Loss]: - target_mask = self.calc_mask(target_tensor) + target_mask = self.calculate_mask(target_tensor) if self.config_mask_factor > 0: target_mask = dilate_mask(target_mask, self.config_mask_factor) @@ -180,7 +180,7 @@ class MaskLoss(nn.Module): weighted_mask_loss = mask_loss * self.config_mask_weight return mask_loss, weighted_mask_loss - def calc_mask(self, target_tensor : Tensor) -> Tensor: + def calculate_mask(self, target_tensor : Tensor) -> Tensor: target_tensor = torch.nn.functional.interpolate(target_tensor, (256, 256), mode = 'bilinear') target_tensor = (target_tensor.clip(-1, 1) + 1) * 0.5 diff --git a/hyperswap/src/training.py b/hyperswap/src/training.py index 592843e..f647ab9 100644 --- a/hyperswap/src/training.py +++ b/hyperswap/src/training.py @@ -14,7 +14,7 @@ from torch.utils.data import ConcatDataset, Dataset, random_split from torchdata.stateful_dataloader import StatefulDataLoader from .dataset import DynamicDataset -from .helper import apply_noise, calc_embedding, erode_mask, overlay_mask +from .helper import apply_noise, calculate_embedding, erode_mask, overlay_mask from .models.discriminator import Discriminator from .models.generator import Generator from .models.loss import AdversarialLoss, CycleLoss, DiscriminatorLoss, FeatureLoss, GazeLoss, IdentityLoss, MaskLoss, ReconstructionLoss @@ -101,8 +101,8 @@ class HyperSwapTrainer(LightningModule): do_update = (batch_index + 1) % self.config_accumulate_size == 0 generator_optimizer, discriminator_optimizer = self.optimizers() #type:ignore[attr-defined] generator_scheduler, discriminator_scheduler = self.lr_schedulers() #type:ignore[attr-defined] - source_embedding = calc_embedding(self.generator_embedder, source_tensor, (0, 0, 0, 0)) - target_embedding = calc_embedding(self.generator_embedder, target_tensor, (0, 0, 0, 0)) + source_embedding = calculate_embedding(self.generator_embedder, source_tensor, (0, 0, 0, 0)) + target_embedding = calculate_embedding(self.generator_embedder, target_tensor, (0, 0, 0, 0)) if self.config_noise_factor > 0: source_embedding = apply_noise(source_embedding, self.config_noise_factor) @@ -176,9 +176,9 @@ class HyperSwapTrainer(LightningModule): def validation_step(self, batch : Batch, batch_index : int) -> Tensor: source_tensor, target_tensor = batch - source_embedding = calc_embedding(self.generator_embedder, source_tensor, (0, 0, 0, 0)) + source_embedding = calculate_embedding(self.generator_embedder, source_tensor, (0, 0, 0, 0)) output_tensor, _ = self.forward(source_embedding, target_tensor) - output_embedding = calc_embedding(self.generator_embedder, output_tensor, (0, 0, 0, 0)) + output_embedding = calculate_embedding(self.generator_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) return validation_score