topoae_2d_std = '/Users/simons/MT_data/eval_data/MNIST_FINAL/TopoAE/std_minimizer/MNIST_offline-seed1988-DeepAE_MNIST-default-lr1_1000-bs64-nep1000-rlw1-tlw1_256-seed1988-1ae25b75'
topoae_2d_cont = '/Users/simons/MT_data/eval_data/MNIST_FINAL/TopoAE/cont_minimizer/MNIST_offline-seed579-DeepAE_MNIST-default-lr1_1000-bs128-nep1000-rlw1-tlw1_128-seed579-3c78a835'
topoae_2d_rec = '/Users/simons/MT_data/eval_data/MNIST_FINAL/TopoAE/rec_minimizer/MNIST_offline-seed838-DeepAE_MNIST-default-lr1_1000-bs256-nep1000-rlw1-tlw1_16-seed838-67de8d97'
for path_exp in [
topoae_2d_kl01, topoae_2d_std, topoae_2d_cont, topoae_2d_rec
]:
# get model
autoencoder = DeepAE_MNIST()
model = TopologicallyRegularizedAutoencoder(autoencoder)
state_dict = torch.load(os.path.join(path_exp, 'model_state.pth'),
map_location=torch.device('cpu'))
model.load_state_dict(state_dict)
model.eval()
dataset = MNIST_offline()
data, labels = dataset.sample(train=False)
z = model.encode(torch.Tensor(data).float())
plot_2Dscatter(z.detach().numpy(),
labels,
path_to_save=os.path.join(
path_exp,
'{}_latent_visualization.pdf'.format('final')),
title=None,
show=False,
palette='custom2')
from src.datasets.datasets import MNIST_offline
from src.evaluation.config import ConfigEval
from src.models.autoencoder.autoencoders import DeepAE_MNIST_4D
from src.models.vanillaAE.config import ConfigGrid_VanillaAE
mnist_1_deepae4 = [
ConfigGrid_VanillaAE(
learning_rate=[1 / 10, 1 / 100, 1 / 1000],
batch_size=[64, 128, 256, 512],
n_epochs=[1000],
weight_decay=[1e-6],
early_stopping=[32],
model_class=[DeepAE_MNIST_4D],
model_kwargs=[dict()],
dataset=[MNIST_offline()],
sampling_kwargs=[
dict(root_path='/cluster/home/schsimo/MT/AEs-VAEs-TDA')
],
eval=[
ConfigEval(active=True,
evaluate_on='test',
eval_manifold=False,
save_eval_latent=False,
save_train_latent=False,
online_visualization=False,
quant_eval=True,
k_min=4,
k_max=16,
k_step=4)
],
import time
import torch
from sklearn.metrics import pairwise_distances
from torch import nn
from src.datasets.datasets import MNIST_offline
data_set = MNIST_offline()
data, labels = data_set.sample(train = True)
data_S = data[:1024,:]
print(data.shape)
start1 = time.time()
distance_matrix1 = pairwise_distances(data, data_S, n_jobs = 2)
end1 = time.time()
print('SKLEARN')
print(end1-start1)
print(distance_matrix1.shape)
print(distance_matrix1[1,0])
print(distance_matrix1[0,1])
data_t = torch.from_numpy(data)
data_ts = torch.from_numpy(data_S)
print(data_t.shape)
start1 = time.time()
distance_matrix3 = torch.cdist(data_t, data_ts)