def __init__(self, ngram, architecture=None, output_size=10):
super(Model, self).__init__()
self.ngram = ngram
self.n = ngram.n
self.cnt = 1
self.output_size = output_size
for idx in ngram:
ngram[idx]
if architecture is None:
self.primal = nn.Sequential(nn.Linear(28 * 28, 300), nn.ReLU(),
nn.Linear(300, 100), nn.ReLU(),
nn.Linear(100, output_size)).to(DEVICE)
else:
self.primal = architecture.to(DEVICE)
self.dual = Ngram(self.ngram.n)
for idx in self.ngram:
self.dual[idx] = torch.tensor(
-1. / self.ngram[idx]).to(DEVICE).requires_grad_()
# self.dual[idx] = torch.tensor(0.).uniform_(-1, 0).to(DEVICE).requires_grad_()
self.to(DEVICE)
self.init_weights()
from config import DEVICE
from src.data_processing import (Ngram, randomized_ngram,
sequence_loader_MNIST, test_loader_MNIST,
train_loader_MNIST)
from src.model import Model
from src.remote import mpl
from src.training import SGD, SPDG
from src.statistics import get_statistics
import matplotlib.pyplot as plt
torch.manual_seed(671470554)
np.random.seed(70914308)
# %% GENERATING DATASET
ngram = Ngram(3)
ngram[(0, 1, 2)] = 6.
ngram[(1, 2, 3)] = 4.
ngram.norm()
data_loader = train_loader_MNIST()
test_loader = test_loader_MNIST()
sequence_loader = sequence_loader_MNIST(ngram, num_samples=40000)
# %% REGULAR TRAINING (SGD)
# model = Model(ngram)
# model.to(DEVICE)
# model.init_weights()
# optimizer = torch.optim.Adam(model.primal.parameters())
# history = SGD(model, optimizer, data_loader, test_loader, num_epochs=1, log_every=50, test_every=1)
# test_every = args.test_every # primal_lr = args.primal_lr # dual_lr = args.dual_lr # else: continuation = False num_epochs = 1000 save_every = 200 log_every = 100 test_every = 5 primal_lr = 1e-6 dual_lr = 1e-4 # %% GENERATING DATASET # ngram = randomized_ngram(3, 2, out_dim=4, min_var=5e-2) ngram = Ngram(3) ngram[(0, 1, 2)] = 9. ngram[(1, 2, 3)] = 1. ngram.norm() ngram.show() # %% CREATING MODEL data_loader = train_loader_MNIST() test_loader = test_loader_MNIST() sequence_loader = sequence_loader_MNIST(ngram, num_samples=50000) sequence_test_loader = sequence_loader_MNIST(ngram, num_samples=10000) # %% REGULAR TRAINING (SGD) # model = Model(ngram) # model.to(DEVICE) # model.init_weights()
from config import DEVICE
from src.data_processing import (Ngram, randomized_ngram,
sequence_loader_MNIST, test_loader_MNIST,
train_loader_MNIST)
from src.model import Model
from src.remote import mpl
from src.training import SGD, SPDG
from src.statistics import get_statistics
import matplotlib.pyplot as plt
torch.manual_seed(377807910)
np.random.seed(713693356)
# %% GENERATING DATASET
ngram = Ngram(1)
ngram[(0)] = 5.
ngram[(1)] = 95.
ngram.norm()
data_loader = train_loader_MNIST()
test_loader = test_loader_MNIST()
sequence_loader = sequence_loader_MNIST(ngram, num_samples=40000)
# %% REGULAR TRAINING (SGD)
# model = Model(ngram)
# model.to(DEVICE)
# model.init_weights()
# optimizer = torch.optim.Adam(model.primal.parameters())
# history = SGD(model, optimizer, data_loader, test_loader, num_epochs=1, log_every=50, test_every=1)
import torch
import numpy as np
import matplotlib as mpl
mpl.use('agg')
import matplotlib.pyplot as plt
from src.data_processing import Ngram, train_loader_MNIST, test_loader_MNIST, sequence_loader_MNIST, randomized_ngram
from src.model import Model
from config import DEVICE
from src.training import SGD, SPDG
torch.manual_seed(403249064)
np.random.seed(235084161)
# %% GENERATING DATASET
n = 1
ngram = Ngram(n)
ngram[(0)] = 1.
ngram[(1)] = 9.
ngram.norm()
data_loader = train_loader_MNIST()
test_loader = test_loader_MNIST()
sequence_loader = sequence_loader_MNIST(ngram, num_samples=20000)
# %% DUAL TRAINING
model = Model(ngram, output_size=2)
model.to(DEVICE)
model.init_weights()
primal_lr = 1e-6
dual_lr = 1e-4