diff --git a/.idea/ae_toolbox_torch.iml b/.idea/ae_toolbox_torch.iml deleted file mode 100644 index 8159b14..0000000 --- a/.idea/ae_toolbox_torch.iml +++ /dev/null @@ -1,16 +0,0 @@ - - - - - - - - - - - - - - \ No newline at end of file diff --git a/.idea/dictionaries/illium.xml b/.idea/dictionaries/illium.xml deleted file mode 100644 index 32b081c..0000000 --- a/.idea/dictionaries/illium.xml +++ /dev/null @@ -1,9 +0,0 @@ - - - - dataloader - datasets - isovists - - - \ No newline at end of file diff --git a/.idea/inspectionProfiles/profiles_settings.xml b/.idea/inspectionProfiles/profiles_settings.xml deleted file mode 100644 index 0eefe32..0000000 --- a/.idea/inspectionProfiles/profiles_settings.xml +++ /dev/null @@ -1,5 +0,0 @@ - - - - \ No newline at end of file diff --git a/.idea/misc.xml b/.idea/misc.xml deleted file mode 100644 index a663f10..0000000 --- a/.idea/misc.xml +++ /dev/null @@ -1,7 +0,0 @@ - - - - - - \ No newline at end of file diff --git a/.idea/modules.xml b/.idea/modules.xml deleted file mode 100644 index fe9fbe4..0000000 --- a/.idea/modules.xml +++ /dev/null @@ -1,8 +0,0 @@ - - - - - - - - \ No newline at end of file diff --git a/.idea/other.xml b/.idea/other.xml deleted file mode 100644 index 640fd80..0000000 --- a/.idea/other.xml +++ /dev/null @@ -1,7 +0,0 @@ - - - - - \ No newline at end of file diff --git a/.idea/vcs.xml b/.idea/vcs.xml deleted file mode 100644 index 94a25f7..0000000 --- a/.idea/vcs.xml +++ /dev/null @@ -1,6 +0,0 @@ - - - - - - \ No newline at end of file diff --git a/.idea/workspace.xml b/.idea/workspace.xml deleted file mode 100644 index 4fa74cf..0000000 --- a/.idea/workspace.xml +++ /dev/null @@ -1,284 +0,0 @@ - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 1564587418949 - - - 1565793753423 - - - 1565958589041 - - - 1565987964760 - - - 1566064016196 - - - 1566366992088 - - - 1566546840536 - - - - - - - - - - - - - - - - - - file://$PROJECT_DIR$/networks/modules.py - 206 - - - file://$PROJECT_DIR$/networks/seperating_adversarial_auto_encoder.py - 23 - - - file://$PROJECT_DIR$/viz/viz_latent.py - 67 - - - - - - - - - - - - - - - - - - - - \ No newline at end of file diff --git a/dataset.py b/dataset.py index a5c85a1..6d160dc 100644 --- a/dataset.py +++ b/dataset.py @@ -167,7 +167,7 @@ class Trajectories(Dataset): dataDict = dict() for key, val in kwargs.items(): if key in self.isovistMeasures: - dataDict[key] = torch.tensor(val) + dataDict[key] = torch.tensor(val, requires_grad=False) # Check if all keys are of same length assert len(set(x.size()[0] for x in dataDict.values() if torch.is_tensor(x))) <= 1 data = torch.stack([dataDict[key] for key in self.isovistMeasures], dim=-1) diff --git a/networks/adverserial_auto_encoder.py b/networks/adverserial_auto_encoder.py index 53352fe..faaeee8 100644 --- a/networks/adverserial_auto_encoder.py +++ b/networks/adverserial_auto_encoder.py @@ -1,7 +1,7 @@ +from torch.optim import Adam + from networks.auto_encoder import AutoEncoder from torch.nn.functional import mse_loss -from torch.nn import Sequential, Linear, ReLU, Dropout, Sigmoid -from torch.distributions import Normal from networks.modules import * import torch @@ -23,14 +23,10 @@ class AdversarialAutoEncoder(AutoEncoder): return z, x_hat -class AdversarialAELightningOverrides: - - @property - def name(self): - return self.__class__.__name__ - - def forward(self, x): - return self.network.forward(x) +class AdversarialAELightningOverrides(LightningModuleOverrides): + + def __init__(self): + super(AdversarialAELightningOverrides, self).__init__() def training_step(self, batch, _, optimizer_i): if optimizer_i == 0: @@ -67,5 +63,12 @@ class AdversarialAELightningOverrides: raise RuntimeError('This should not have happened, catch me if u can.') + # This is Fucked up, why do i need to put an additional empty list here? + def configure_optimizers(self): + return [Adam(self.network.discriminator.parameters(), lr=0.02), + Adam([*self.network.encoder.parameters(), *self.network.decoder.parameters()], lr=0.02)],\ + [] + + if __name__ == '__main__': raise PermissionError('Get out of here - never run this module') diff --git a/networks/auto_encoder.py b/networks/auto_encoder.py index b72bc59..a834d1d 100644 --- a/networks/auto_encoder.py +++ b/networks/auto_encoder.py @@ -1,3 +1,5 @@ +from torch.optim import Adam + from .modules import * from torch.nn.functional import mse_loss from torch import Tensor @@ -26,14 +28,10 @@ class AutoEncoder(AbstractNeuralNetwork, ABC): return z, x_hat -class AutoEncoderLightningOverrides: +class AutoEncoderLightningOverrides(LightningModuleOverrides): - @property - def name(self): - return self.__class__.__name__ - - def forward(self, x): - return self.network.forward(x) + def __init__(self): + super(AutoEncoderLightningOverrides, self).__init__() def training_step(self, x, batch_nb): # z, x_hat @@ -41,6 +39,9 @@ class AutoEncoderLightningOverrides: loss = mse_loss(x, x_hat) return {'loss': loss} + def configure_optimizers(self): + return [Adam(self.parameters(), lr=0.02)] + if __name__ == '__main__': raise PermissionError('Get out of here - never run this module') diff --git a/networks/modules.py b/networks/modules.py index 0cc5ccf..81a02bb 100644 --- a/networks/modules.py +++ b/networks/modules.py @@ -1,11 +1,34 @@ +import os + import torch import pytorch_lightning as pl -from torch.nn import Module, Linear, ReLU, Tanh, Sigmoid, Dropout, GRU, AvgPool2d +from pytorch_lightning import data_loader +from torch.nn import Module, Linear, ReLU, Tanh, Sigmoid, Dropout, GRU from abc import ABC, abstractmethod ####################### -# Abstract NN Class +# Abstract NN Class & Lightning Module +from torch.utils.data import DataLoader + +from dataset import DataContainer + + +class LightningModuleOverrides: + + @property + def name(self): + return self.__class__.__name__ + + def forward(self, x): + return self.network.forward(x) + + @data_loader + def tng_dataloader(self): + num_workers = os.cpu_count() // 2 + return DataLoader(DataContainer('data', self.size, self.step), + shuffle=True, batch_size=100, num_workers=num_workers) + class AbstractNeuralNetwork(Module): diff --git a/networks/seperating_adversarial_auto_encoder.py b/networks/seperating_adversarial_auto_encoder.py index 5bf5fc5..b0872f4 100644 --- a/networks/seperating_adversarial_auto_encoder.py +++ b/networks/seperating_adversarial_auto_encoder.py @@ -1,3 +1,5 @@ +from torch.optim import Adam + from networks.auto_encoder import AutoEncoder from torch.nn.functional import mse_loss from networks.modules import * @@ -7,16 +9,15 @@ import torch class SeperatingAdversarialAutoEncoder(Module): def __init__(self, latent_dim, features, **kwargs): - assert latent_dim % 2 == 0, f'Your latent space needs to be even, not odd, but was: "{latent_dim}"' super(SeperatingAdversarialAutoEncoder, self).__init__() self.latent_dim = latent_dim self.features = features - self.spatial_encoder = PoolingEncoder(self.latent_dim // 2) - self.temporal_encoder = Encoder(self.latent_dim // 2) + self.spatial_encoder = PoolingEncoder(self.latent_dim) + self.temporal_encoder = Encoder(self.latent_dim) self.decoder = Decoder(self.latent_dim, self.features) - self.spatial_discriminator = Discriminator(self.latent_dim // 2, self.features) - self.temporal_discriminator = Discriminator(self.latent_dim // 2, self.features) + self.spatial_discriminator = Discriminator(self.latent_dim, self.features) + self.temporal_discriminator = Discriminator(self.latent_dim, self.features) def forward(self, batch): # Encoder @@ -30,14 +31,10 @@ class SeperatingAdversarialAutoEncoder(Module): return z_spatial, z_temporal, x_hat -class SeparatingAdversarialAELightningOverrides: +class SeparatingAdversarialAELightningOverrides(LightningModuleOverrides): - @property - def name(self): - return self.__class__.__name__ - - def forward(self, x): - return self.network.forward(x) + def __init__(self): + super(SeparatingAdversarialAELightningOverrides, self).__init__() def training_step(self, batch, _, optimizer_i): spatial_latent_fake, temporal_latent_fake, batch_hat = self.network.forward(batch) @@ -91,6 +88,17 @@ class SeparatingAdversarialAELightningOverrides: else: raise RuntimeError('This should not have happened, catch me if u can.') + # This is Fucked up, why do i need to put an additional empty list here? + def configure_optimizers(self): + return [Adam([*self.network.spatial_discriminator.parameters(), *self.network.spatial_encoder.parameters()] + , lr=0.02), + Adam([*self.network.temporal_discriminator.parameters(), *self.network.temporal_encoder.parameters()] + , lr=0.02), + Adam([*self.network.temporal_encoder.parameters(), + *self.network.spatial_encoder.parameters(), + *self.network.decoder.parameters()] + , lr=0.02)], [] + if __name__ == '__main__': raise PermissionError('Get out of here - never run this module') diff --git a/networks/variational_auto_encoder.py b/networks/variational_auto_encoder.py index 64cb7a9..aad4a54 100644 --- a/networks/variational_auto_encoder.py +++ b/networks/variational_auto_encoder.py @@ -1,3 +1,5 @@ +from torch.optim import Adam + from .modules import * from torch.nn.functional import mse_loss @@ -33,14 +35,10 @@ class VariationalAutoEncoder(AbstractNeuralNetwork, ABC): return x_hat, mu, logvar -class VariationalAutoEncoderLightningOverrides: +class VariationalAutoEncoderLightningOverrides(LightningModuleOverrides): - @property - def name(self): - return self.network.name - - def forward(self, x): - return self.network.forward(x) + def __init__(self): + super(VariationalAutoEncoderLightningOverrides, self).__init__() def training_step(self, x, _): x_hat, logvar, mu = self.forward(x) @@ -53,6 +51,9 @@ class VariationalAutoEncoderLightningOverrides: KLD = -0.5 * torch.sum(1 + logvar - mu.pow(2) - logvar.exp()) return {'loss': BCE + KLD} + def configure_optimizers(self): + return [Adam(self.parameters(), lr=0.02)] + if __name__ == '__main__': raise PermissionError('Get out of here - never run this module') diff --git a/run_models.py b/run_models.py index 6362036..228430d 100644 --- a/run_models.py +++ b/run_models.py @@ -1,3 +1,5 @@ +from torch.distributions import Normal + from networks.auto_encoder import * import os import time @@ -18,90 +20,54 @@ from argparse import Namespace from argparse import ArgumentParser args = ArgumentParser() -args.add_argument('step') -args.add_argument('features') -args.add_argument('size') -args.add_argument('latent_dim') +args.add_argument('--step', default=0) +args.add_argument('--features', default=0) +args.add_argument('--size', default=0) +args.add_argument('--latent_dim', default=0) +args.add_argument('--model', default='Model') # ToDo: How to implement this better? # other_classes = [AutoEncoder, AutoEncoderLightningOverrides] class Model(AutoEncoderLightningOverrides, LightningModule): - def __init__(self, latent_dim=0, size=0, step=0, features=0, **kwargs): - assert all([x in args for x in ['step', 'size', 'latent_dim', 'features']]) - self.size = args.size - self.latent_dim = args.latent_dim - self.features = args.features - self.step = args.step + def __init__(self, parameters, **kwargs): + assert all([x in parameters for x in ['step', 'size', 'latent_dim', 'features']]) + self.size = parameters.size + self.latent_dim = parameters.latent_dim + self.features = parameters.features + self.step = parameters.step super(Model, self).__init__() self.network = AutoEncoder(self.latent_dim, self.features) - def configure_optimizers(self): - return [Adam(self.parameters(), lr=0.02)] - - @data_loader - def tng_dataloader(self): - return DataLoader(DataContainer('data', self.size, self.step), shuffle=True, batch_size=100) - class AdversarialModel(AdversarialAELightningOverrides, LightningModule): - @property - def name(self): - return self.network.name - - def __init__(self, args: Namespace, **kwargs): - assert all([x in args for x in ['step', 'size', 'latent_dim', 'features']]) - self.size = args.size - self.latent_dim = args.latent_dim - self.features = args.features - self.step = args.step + def __init__(self, parameters: Namespace, **kwargs): + assert all([x in parameters for x in ['step', 'size', 'latent_dim', 'features']]) + self.size = parameters.size + self.latent_dim = parameters.latent_dim + self.features = parameters.features + self.step = parameters.step super(AdversarialModel, self).__init__() self.normal = Normal(0, 1) self.network = AdversarialAutoEncoder(self.latent_dim, self.features) pass - # This is Fucked up, why do i need to put an additional empty list here? - def configure_optimizers(self): - return [Adam(self.network.discriminator.parameters(), lr=0.02), - Adam([*self.network.encoder.parameters(), *self.network.decoder.parameters()], lr=0.02)],\ - [] - - @data_loader - def tng_dataloader(self): - return DataLoader(DataContainer('data', self.size, self.step), shuffle=True, batch_size=100) - class SeparatingAdversarialModel(SeparatingAdversarialAELightningOverrides, LightningModule): - def __init__(self, args: Namespace, **kwargs): - assert all([x in args for x in ['step', 'size', 'latent_dim', 'features']]) - self.size = args.size - self.latent_dim = args.latent_dim - self.features = args.features - self.step = args.step + def __init__(self, parameters: Namespace, **kwargs): + assert all([x in parameters for x in ['step', 'size', 'latent_dim', 'features']]) + self.size = parameters.size + self.latent_dim = parameters.latent_dim + self.features = parameters.features + self.step = parameters.step super(SeparatingAdversarialModel, self).__init__() self.normal = Normal(0, 1) self.network = SeperatingAdversarialAutoEncoder(self.latent_dim, self.features, **kwargs) pass - # This is Fucked up, why do i need to put an additional empty list here? - def configure_optimizers(self): - return [Adam([*self.network.spatial_discriminator.parameters(), *self.network.spatial_encoder.parameters()] - , lr=0.02), - Adam([*self.network.temporal_discriminator.parameters(), *self.network.temporal_encoder.parameters()] - , lr=0.02), - Adam([*self.network.temporal_encoder.parameters(), - *self.network.spatial_encoder.parameters(), - *self.network.decoder.parameters()] - , lr=0.02)], [] - - @data_loader - def tng_dataloader(self): - num_workers = os.cpu_count() // 2 - return DataLoader(DataContainer('data', self.size, self.step), shuffle=True, batch_size=100, num_workers=num_workers) - if __name__ == '__main__': features = 6 @@ -110,7 +76,7 @@ if __name__ == '__main__': arguments = args.parse_args() arguments.__dict__.update(tag_dict) - model = SeparatingAdversarialModel(arguments) + model = globals()[arguments.model](arguments) # PyTorch summarywriter with a few bells and whistles outpath = os.path.join(os.getcwd(), 'output', model.name, time.asctime().replace(' ', '_').replace(':', '-')) @@ -124,7 +90,7 @@ if __name__ == '__main__': filepath=os.path.join(outpath, 'weights.ckpt'), save_best_only=True, verbose=True, - monitor='tng_loss', # val_loss + monitor='val_loss', # val_loss mode='min', ) diff --git a/viz/viz_latent.py b/viz/viz_latent.py index 140304c..15e61f0 100644 --- a/viz/viz_latent.py +++ b/viz/viz_latent.py @@ -1,21 +1,17 @@ -# TODO: THIS -import seaborn as sb -import torch -from torch.utils.data import DataLoader -from pytorch_lightning import data_loader -from dataset import DataContainer from collections import defaultdict from tqdm import tqdm -import os from sklearn.manifold import TSNE from sklearn.decomposition import PCA -import seaborn as sns; sns.set() +import seaborn as sns import matplotlib.pyplot as plt from run_models import * +sns.set() + + def search_for_weights(folder): while not os.path.exists(folder): if len(os.path.split(folder)) >= 50: @@ -32,6 +28,8 @@ def search_for_weights(folder): def load_and_predict(path_like_element): + if any([x.name.endswith('.png') for x in os.scandir(os.path.dirname(path_like_element))]): + return # Define Loop to search for models and folder with visualizations model = globals()[path_like_element.path.split(os.sep)[-3]] @@ -46,36 +44,50 @@ def load_and_predict(path_like_element): pretrained_model.eval() pretrained_model.freeze() - # Load the data for prediction - dataset = DataContainer(os.path.join(os.pardir, 'data'), 5, 5) + with torch.no_grad(): - # Do the inference - prediction_dict = defaultdict(list) - for i in tqdm(range(len(dataset)), total=len(dataset)): - p_X = pretrained_model(dataset[i].unsqueeze(0)) - for idx in range(len(p_X) - 1): - prediction_dict[idx].append(p_X[idx]) + # Load the data for prediction + dataset = DataContainer(os.path.join(os.pardir, 'data'), 5, 5) + + # Do the inference + prediction_dict = defaultdict(list) + for i in tqdm(range(len(dataset)), total=len(dataset)): + p_X = pretrained_model(dataset[i].unsqueeze(0)) + for idx in range(len(p_X) - 1): + prediction_dict[idx].append(p_X[idx]) predictions = [torch.cat(prediction).detach().numpy() for prediction in prediction_dict.values()] - for prediction in predictions: - viz_latent(prediction) + for idx, prediction in enumerate(predictions): + plot, _ = viz_latent(prediction) + plot.savefig(os.path.join(os.path.dirname(path_like_element), f'latent_space_{idx}.png')) -def viz_latent(prediction): +def viz_latent(prediction, title=f'Latent Space '): if prediction.shape[-1] <= 1: raise ValueError('How did this happen?') elif prediction.shape[-1] == 2: ax = sns.scatterplot(x=prediction[:, 0], y=prediction[:, 1]) - plt.show() - return ax + try: + plt.show() + except: + pass + return ax.figure, (ax) else: fig, axs = plt.subplots(ncols=2) - predictions_pca = PCA(n_components=2) - predictions_tsne = TSNE(n_components=2) - pca_plot = sns.scatterplot(x=predictions_pca[:, 0], y=predictions_pca[:, 1], ax=axs[0]) - tsne_plot = sns.scatterplot(x=predictions_tsne[:, 0], y=predictions_tsne[:, 1], ax=axs[1]) - plt.show() - return fig, axs, pca_plot, tsne_plot + plots = [] + for idx, dim_reducer in enumerate([PCA, TSNE]): + predictions_reduced = dim_reducer(n_components=2).fit_transform(prediction) + plot = sns.scatterplot(x=predictions_reduced[:, 0], y=predictions_reduced[:, 1], + ax=axs[idx]) + plot.set_title(dim_reducer.__name__) + plots.append(plot) + + try: + plt.show() + except: + pass + return fig, (*plots, ) + if __name__ == '__main__': path = 'output'