Linear Regression using PyTorch¶
In [1]:
import matplotlib.pyplot as plt
%matplotlib inline
import numpy as np
In [2]:
N = 10 # number of data points
m = .7
c = 0
x = np.linspace(0,2*np.pi,N)
y = m*x + c + np.random.normal(0,.3,x.shape)
plt.figure()
plt.plot(x,y,'o')
plt.xlabel('x')
plt.ylabel('y')
plt.title('2D data (#data = %d)' % N)
plt.show()
In [3]:
import torch
Dataset¶
In [4]:
from torch.utils.data import Dataset
class MyDataset(Dataset):
def __init__(self, x, y):
self.x = x
self.y = y
def __len__(self):
return len(self.x)
def __getitem__(self, idx):
sample = {
'feature': torch.tensor([1,self.x[idx]]),
'label': torch.tensor([self.y[idx]])}
return sample
In [5]:
dataset = MyDataset(x, y)
for i in range(len(dataset)):
sample = dataset[i]
print(i, sample['feature'], sample['label'])
0 tensor([1., 0.]) tensor([0.1019]) 1 tensor([1.0000, 0.6981]) tensor([0.3657]) 2 tensor([1.0000, 1.3963]) tensor([0.7600]) 3 tensor([1.0000, 2.0944]) tensor([1.5153]) 4 tensor([1.0000, 2.7925]) tensor([2.3835]) 5 tensor([1.0000, 3.4907]) tensor([3.0468]) 6 tensor([1.0000, 4.1888]) tensor([2.5200]) 7 tensor([1.0000, 4.8869]) tensor([3.6476]) 8 tensor([1.0000, 5.5851]) tensor([3.9853]) 9 tensor([1.0000, 6.2832]) tensor([4.6717])
Dataloader¶
In [6]:
from torch.utils.data import DataLoader
dataset = MyDataset(x, y)
batch_size = 4
shuffle = True
num_workers = 4
dataloader = DataLoader(dataset, batch_size=batch_size, shuffle=shuffle, num_workers=num_workers)
In [7]:
import pprint as pp
for i_batch, samples in enumerate(dataloader):
print('\nbatch# = %s' % i_batch)
print('samples: ')
pp.pprint(samples)
batch# = 0
samples:
{'feature': tensor([[1.0000, 2.0944],
[1.0000, 6.2832],
[1.0000, 1.3963],
[1.0000, 4.1888]]),
'label': tensor([[1.5153],
[4.6717],
[0.7600],
[2.5200]])}
batch# = 1
samples:
{'feature': tensor([[1.0000, 5.5851],
[1.0000, 2.7925],
[1.0000, 0.0000],
[1.0000, 3.4907]]),
'label': tensor([[3.9853],
[2.3835],
[0.1019],
[3.0468]])}
batch# = 2
samples:
{'feature': tensor([[1.0000, 4.8869],
[1.0000, 0.6981]]),
'label': tensor([[3.6476],
[0.3657]])}
Model¶
In [8]:
import torch.nn as nn
from torch.nn.parameter import Parameter
class MyModel(nn.Module):
def __init__(self, input_dim, output_dim):
super(MyModel, self).__init__()
self.weight = Parameter(torch.Tensor(output_dim, input_dim))
self.bias = Parameter(torch.Tensor(output_dim, 1))
stdv = 1.
self.weight.data.uniform_(-stdv, stdv)
self.bias.data.uniform_(-stdv, stdv)
def forward(self, x):
weight_and_bias = torch.cat((self.weight, self.bias), 1)
#print(weight_and_bias)
#print(weight_and_bias.t().shape)
#print(x.shape)
#print(self.weight.size())
out = x.matmul(weight_and_bias.t())
return out
Setting a model for our problem¶
In [9]:
input_dim = 1
output_dim = 1
model = MyModel(input_dim, output_dim)
#model(torch.rand([5,2]))
Cost function¶
Often called loss or error
In [10]:
#cost = nn.MSELoss()
In [11]:
import torch.nn as nn
class MyLoss(nn.Module):
def __init__(self):
super(MyLoss, self).__init__()
def forward(self, predictions, targets):
print(predictions)
print(targets)
diff = torch.sub(predictions, targets)
diff2 = torch.pow(diff, 2)
err = torch.sum(diff2)
return err
In [12]:
cost = MyLoss()
Minimizing the cost function¶
In other words training (or learning from data)
In [13]:
num_epochs = 100 # How many times the entire training data is seen?
l_rate = 0.01
optimiser = torch.optim.SGD(model.parameters(), lr = l_rate)
dataset = MyDataset(x, y)
batch_size = 4
shuffle = True
num_workers = 4
training_sample_generator = DataLoader(dataset, batch_size=batch_size, shuffle=shuffle, num_workers=num_workers)
for epoch in range(num_epochs):
if epoch % 10 == 0:
print('Epoch = %s' % epoch)
for batch_i, samples in enumerate(training_sample_generator):
predictions = model(samples['feature'])
error = cost(predictions, samples['label'])
if epoch % 10 == 0:
print('\tBatch = %s, Error = %s' % (batch_i, error.item()))
# Before the backward pass, use the optimizer object to zero all of the
# gradients for the variables it will update (which are the learnable
# weights of the model). This is because by default, gradients are
# accumulated in buffers( i.e, not overwritten) whenever .backward()
# is called. Checkout docs of torch.autograd.backward for more details.
optimiser.zero_grad()
# Backward pass: compute gradient of the loss with respect to model
# parameters
error.backward()
# Calling the step function on an Optimizer makes an update to its
# parameters
optimiser.step()
Epoch = 0
tensor([[2.1582],
[1.2339],
[1.0491],
[1.7885]], grad_fn=<MmBackward>)
tensor([[3.9853],
[1.5153],
[0.7600],
[2.5200]])
Batch = 0, Error = 4.036039352416992
tensor([[0.7303],
[1.1031],
[3.3394],
[2.2212]], grad_fn=<MmBackward>)
tensor([[0.1019],
[0.3657],
[3.6476],
[2.3835]])
Batch = 1, Error = 1.0599795579910278
tensor([[4.2485],
[2.6769]], grad_fn=<MmBackward>)
tensor([[4.6717],
[3.0468]])
Batch = 2, Error = 0.316006064414978
tensor([[2.9686],
[1.1764],
[4.3127],
[3.8647]], grad_fn=<MmBackward>)
tensor([[3.0468],
[0.3657],
[3.9853],
[3.6476]])
tensor([[3.1245],
[2.3173],
[4.3354],
[0.7028]], grad_fn=<MmBackward>)
tensor([[2.5200],
[2.3835],
[4.6717],
[0.1019]])
tensor([[1.4874],
[1.8878]], grad_fn=<MmBackward>)
tensor([[0.7600],
[1.5153]])
tensor([[2.1658],
[4.0423],
[2.5411],
[3.6670]], grad_fn=<MmBackward>)
tensor([[2.3835],
[4.6717],
[3.0468],
[3.9853]])
tensor([[4.1174],
[2.1635],
[1.1866],
[3.6289]], grad_fn=<MmBackward>)
tensor([[3.6476],
[1.5153],
[0.3657],
[2.5200]])
tensor([[0.6372],
[1.3664]], grad_fn=<MmBackward>)
tensor([[0.1019],
[0.7600]])
tensor([[1.3199],
[0.9671],
[3.0838],
[1.6727]], grad_fn=<MmBackward>)
tensor([[0.7600],
[0.3657],
[3.6476],
[1.5153]])
tensor([[2.8184],
[2.4486],
[0.5992],
[3.9281]], grad_fn=<MmBackward>)
tensor([[2.5200],
[3.0468],
[0.1019],
[4.6717]])
tensor([[4.1847],
[2.3974]], grad_fn=<MmBackward>)
tensor([[3.9853],
[2.3835]])
tensor([[2.7595],
[0.6059],
[1.0366],
[4.4824]], grad_fn=<MmBackward>)
tensor([[3.0468],
[0.1019],
[0.3657],
[4.6717]])
tensor([[1.9563],
[4.2303],
[3.7755],
[1.5015]], grad_fn=<MmBackward>)
tensor([[1.5153],
[3.9853],
[3.6476],
[0.7600]])
tensor([[2.1593],
[2.9585]], grad_fn=<MmBackward>)
tensor([[2.3835],
[2.5200]])
tensor([[0.5565],
[3.2355],
[3.6182],
[2.4701]], grad_fn=<MmBackward>)
tensor([[0.1019],
[3.6476],
[3.9853],
[3.0468]])
tensor([[3.3800],
[1.0421],
[1.5097],
[1.9773]], grad_fn=<MmBackward>)
tensor([[2.5200],
[0.3657],
[0.7600],
[1.5153]])
tensor([[2.0498],
[3.9626]], grad_fn=<MmBackward>)
tensor([[2.3835],
[4.6717]])
tensor([[0.9982],
[2.3716],
[1.9138],
[1.4560]], grad_fn=<MmBackward>)
tensor([[0.3657],
[2.3835],
[1.5153],
[0.7600]])
tensor([[4.3478],
[3.4941],
[3.0673],
[2.6404]], grad_fn=<MmBackward>)
tensor([[4.6717],
[3.6476],
[2.5200],
[3.0468]])
tensor([[4.1412],
[0.5129]], grad_fn=<MmBackward>)
tensor([[3.9853],
[0.1019]])
tensor([[0.5016],
[0.9429],
[3.1499],
[2.7085]], grad_fn=<MmBackward>)
tensor([[0.1019],
[0.3657],
[2.5200],
[3.0468]])
tensor([[1.3070],
[2.1378],
[1.7224],
[3.3840]], grad_fn=<MmBackward>)
tensor([[0.7600],
[2.3835],
[1.5153],
[3.6476]])
tensor([[3.8813],
[4.3075]], grad_fn=<MmBackward>)
tensor([[3.9853],
[4.6717]])
tensor([[4.2112],
[2.8122],
[4.6775],
[1.8796]], grad_fn=<MmBackward>)
tensor([[3.9853],
[3.0468],
[4.6717],
[1.5153]])
tensor([[1.3714],
[0.4734],
[2.2693],
[3.1673]], grad_fn=<MmBackward>)
tensor([[0.7600],
[0.1019],
[2.3835],
[2.5200]])
tensor([[0.8468],
[3.2687]], grad_fn=<MmBackward>)
tensor([[0.3657],
[3.6476]])
tensor([[1.2907],
[2.5652],
[0.8659],
[1.7155]], grad_fn=<MmBackward>)
tensor([[0.7600],
[3.0468],
[0.3657],
[1.5153]])
tensor([[0.4261],
[4.2710],
[2.9894],
[3.4166]], grad_fn=<MmBackward>)
tensor([[0.1019],
[4.6717],
[2.5200],
[3.6476]])
tensor([[4.0283],
[2.2256]], grad_fn=<MmBackward>)
tensor([[3.9853],
[2.3835]])
Epoch = 10
tensor([[3.5996],
[2.2391],
[1.3321],
[4.5066]], grad_fn=<MmBackward>)
tensor([[3.6476],
[2.3835],
[0.7600],
[4.6717]])
Batch = 0, Error = 0.37773826718330383
tensor([[0.4208],
[0.8866],
[1.8181],
[4.1467]], grad_fn=<MmBackward>)
tensor([[0.1019],
[0.3657],
[1.5153],
[3.9853]])
Batch = 1, Error = 0.4907659888267517
tensor([[2.5909],
[3.0301]], grad_fn=<MmBackward>)
tensor([[3.0468],
[2.5200]])
Batch = 2, Error = 0.46807536482810974
tensor([[1.2569],
[3.4149],
[0.3937],
[2.9833]], grad_fn=<MmBackward>)
tensor([[0.7600],
[3.6476],
[0.1019],
[2.5200]])
tensor([[4.0696],
[0.7840],
[1.6054],
[2.0161]], grad_fn=<MmBackward>)
tensor([[4.6717],
[0.3657],
[1.5153],
[2.3835]])
tensor([[2.7382],
[4.1516]], grad_fn=<MmBackward>)
tensor([[3.0468],
[3.9853]])
tensor([[3.2246],
[4.1711],
[1.3318],
[0.8586]], grad_fn=<MmBackward>)
tensor([[2.5200],
[3.9853],
[0.7600],
[0.3657]])
tensor([[3.1572],
[0.3463],
[1.9525],
[1.5510]], grad_fn=<MmBackward>)
tensor([[3.6476],
[0.1019],
[2.3835],
[1.5153]])
tensor([[2.6130],
[4.4162]], grad_fn=<MmBackward>)
tensor([[3.0468],
[4.6717]])
tensor([[0.8672],
[0.3729],
[1.3616],
[4.8220]], grad_fn=<MmBackward>)
tensor([[0.3657],
[0.1019],
[0.7600],
[4.6717]])
tensor([[2.6651],
[3.5942],
[4.0588],
[3.1297]], grad_fn=<MmBackward>)
tensor([[3.0468],
[3.6476],
[3.9853],
[2.5200]])
tensor([[2.1191],
[1.6737]], grad_fn=<MmBackward>)
tensor([[2.3835],
[1.5153]])
tensor([[3.0461],
[2.1439],
[1.6928],
[3.4972]], grad_fn=<MmBackward>)
tensor([[2.5200],
[2.3835],
[1.5153],
[3.6476]])
tensor([[2.5070],
[0.7680],
[4.2459],
[0.3333]], grad_fn=<MmBackward>)
tensor([[3.0468],
[0.3657],
[4.6717],
[0.1019]])
tensor([[1.3289],
[4.2958]], grad_fn=<MmBackward>)
tensor([[0.7600],
[3.9853]])
tensor([[3.0774],
[3.5366],
[1.2407],
[1.6999]], grad_fn=<MmBackward>)
tensor([[2.5200],
[3.6476],
[0.7600],
[1.5153]])
tensor([[4.0746],
[0.7195],
[3.6552],
[1.9776]], grad_fn=<MmBackward>)
tensor([[4.6717],
[0.3657],
[3.9853],
[2.3835]])
tensor([[0.3197],
[2.8692]], grad_fn=<MmBackward>)
tensor([[0.1019],
[3.0468]])
tensor([[0.3189],
[3.4302],
[1.8746],
[1.3560]], grad_fn=<MmBackward>)
tensor([[0.1019],
[2.5200],
[1.5153],
[0.7600]])
tensor([[2.4932],
[3.8228],
[4.2660],
[2.0500]], grad_fn=<MmBackward>)
tensor([[3.0468],
[3.9853],
[4.6717],
[2.3835]])
tensor([[4.0265],
[0.8378]], grad_fn=<MmBackward>)
tensor([[3.6476],
[0.3657]])
tensor([[4.2973],
[2.7943],
[0.2893],
[3.7963]], grad_fn=<MmBackward>)
tensor([[3.9853],
[3.0468],
[0.1019],
[3.6476]])
tensor([[1.2391],
[4.5909],
[1.7179],
[3.1544]], grad_fn=<MmBackward>)
tensor([[0.7600],
[4.6717],
[1.5153],
[2.5200]])
tensor([[0.6903],
[1.9909]], grad_fn=<MmBackward>)
tensor([[0.3657],
[2.3835]])
tensor([[2.9323],
[3.8237],
[1.1495],
[0.2581]], grad_fn=<MmBackward>)
tensor([[2.5200],
[3.9853],
[0.7600],
[0.1019]])
tensor([[1.9486],
[0.6687],
[1.5220],
[3.2284]], grad_fn=<MmBackward>)
tensor([[2.3835],
[0.3657],
[1.5153],
[3.6476]])
tensor([[4.4742],
[2.5981]], grad_fn=<MmBackward>)
tensor([[4.6717],
[3.0468]])
tensor([[1.7906],
[1.2824],
[2.8070],
[3.8235]], grad_fn=<MmBackward>)
tensor([[1.5153],
[0.7600],
[3.0468],
[3.6476]])
tensor([[4.6583],
[0.7409],
[0.2513],
[2.2099]], grad_fn=<MmBackward>)
tensor([[4.6717],
[0.3657],
[0.1019],
[2.3835]])
tensor([[3.2083],
[4.1962]], grad_fn=<MmBackward>)
tensor([[2.5200],
[3.9853]])
Epoch = 20
tensor([[2.8501],
[1.5383],
[1.9756],
[4.1619]], grad_fn=<MmBackward>)
tensor([[2.5200],
[1.5153],
[2.3835],
[4.6717]])
Batch = 0, Error = 0.5358420014381409
tensor([[0.7157],
[0.2378],
[1.1937],
[4.0611]], grad_fn=<MmBackward>)
tensor([[0.3657],
[0.1019],
[0.7600],
[3.9853]])
Batch = 1, Error = 0.33479544520378113
tensor([[3.4389],
[2.5186]], grad_fn=<MmBackward>)
tensor([[3.6476],
[3.0468]])
Batch = 2, Error = 0.32255789637565613
tensor([[3.7335],
[2.7333],
[0.2327],
[4.2336]], grad_fn=<MmBackward>)
tensor([[3.6476],
[3.0468],
[0.1019],
[3.9853]])
tensor([[1.7002],
[1.2100],
[2.1903],
[0.7198]], grad_fn=<MmBackward>)
tensor([[1.5153],
[0.7600],
[2.3835],
[0.3657]])
tensor([[4.5344],
[3.0942]], grad_fn=<MmBackward>)
tensor([[4.6717],
[2.5200]])
tensor([[3.4148],
[4.3319],
[1.1221],
[0.6635]], grad_fn=<MmBackward>)
tensor([[3.6476],
[4.6717],
[0.7600],
[0.3657]])
tensor([[0.2032],
[3.1690],
[4.1575],
[2.6747]], grad_fn=<MmBackward>)
tensor([[0.1019],
[2.5200],
[3.9853],
[3.0468]])
tensor([[1.5754],
[2.0364]], grad_fn=<MmBackward>)
tensor([[1.5153],
[2.3835]])
tensor([[0.6708],
[3.5077],
[3.0349],
[1.1436]], grad_fn=<MmBackward>)
tensor([[0.3657],
[3.6476],
[2.5200],
[0.7600]])
tensor([[1.9439],
[3.7111],
[1.5021],
[2.3857]], grad_fn=<MmBackward>)
tensor([[2.3835],
[3.9853],
[1.5153],
[3.0468]])
tensor([[0.2045],
[4.8209]], grad_fn=<MmBackward>)
tensor([[0.1019],
[4.6717]])
tensor([[1.6990],
[4.6981],
[0.6993],
[4.1982]], grad_fn=<MmBackward>)
tensor([[1.5153],
[4.6717],
[0.3657],
[3.9853]])
tensor([[3.0182],
[3.4905],
[0.1843],
[2.0736]], grad_fn=<MmBackward>)
tensor([[2.5200],
[3.6476],
[0.1019],
[2.3835]])
tensor([[1.1140],
[2.5119]], grad_fn=<MmBackward>)
tensor([[0.7600],
[3.0468]])
tensor([[2.6114],
[3.0965],
[0.6708],
[0.1856]], grad_fn=<MmBackward>)
tensor([[3.0468],
[2.5200],
[0.3657],
[0.1019]])
tensor([[2.0538],
[3.4628],
[4.4021],
[1.5841]], grad_fn=<MmBackward>)
tensor([[2.3835],
[3.6476],
[4.6717],
[1.5153]])
tensor([[4.3236],
[1.2229]], grad_fn=<MmBackward>)
tensor([[3.9853],
[0.7600]])
tensor([[2.0988],
[0.6547],
[1.1361],
[4.0244]], grad_fn=<MmBackward>)
tensor([[2.3835],
[0.3657],
[0.7600],
[3.9853]])
tensor([[1.6028],
[3.0407],
[4.4785],
[0.1649]], grad_fn=<MmBackward>)
tensor([[1.5153],
[2.5200],
[4.6717],
[0.1019]])
tensor([[3.3980],
[2.4715]], grad_fn=<MmBackward>)
tensor([[3.6476],
[3.0468]])
tensor([[4.2383],
[2.7134],
[2.2051],
[1.1885]], grad_fn=<MmBackward>)
tensor([[3.9853],
[3.0468],
[2.3835],
[0.7600]])
tensor([[0.6719],
[0.1685],
[1.6788],
[4.6994]], grad_fn=<MmBackward>)
tensor([[0.3657],
[0.1019],
[1.5153],
[4.6717]])
tensor([[3.6099],
[3.1167]], grad_fn=<MmBackward>)
tensor([[3.6476],
[2.5200]])
tensor([[0.1460],
[2.4506],
[2.9115],
[1.9897]], grad_fn=<MmBackward>)
tensor([[0.1019],
[3.0468],
[2.5200],
[2.3835]])
tensor([[3.5341],
[1.1220],
[1.6044],
[4.0166]], grad_fn=<MmBackward>)
tensor([[3.6476],
[0.7600],
[1.5153],
[3.9853]])
tensor([[4.4524],
[0.6278]], grad_fn=<MmBackward>)
tensor([[4.6717],
[0.3657]])
tensor([[4.1070],
[0.6436],
[4.6017],
[2.1279]], grad_fn=<MmBackward>)
tensor([[3.9853],
[0.3657],
[4.6717],
[2.3835]])
tensor([[1.6434],
[0.1474],
[3.1394],
[1.1447]], grad_fn=<MmBackward>)
tensor([[1.5153],
[0.1019],
[2.5200],
[0.7600]])
tensor([[3.2822],
[2.3798]], grad_fn=<MmBackward>)
tensor([[3.6476],
[3.0468]])
Epoch = 30
tensor([[0.1445],
[4.2136],
[1.6704],
[0.6531]], grad_fn=<MmBackward>)
tensor([[0.1019],
[3.9853],
[1.5153],
[0.3657]])
Batch = 0, Error = 0.16057457029819489
tensor([[3.5147],
[2.0642],
[1.0972],
[2.5477]], grad_fn=<MmBackward>)
tensor([[3.6476],
[2.3835],
[0.7600],
[3.0468]])
Batch = 1, Error = 0.48238328099250793
tensor([[4.8474],
[3.2791]], grad_fn=<MmBackward>)
tensor([[4.6717],
[2.5200]])
Batch = 2, Error = 0.6071034669876099
tensor([[3.3645],
[1.5127],
[4.2904],
[1.0497]], grad_fn=<MmBackward>)
tensor([[3.6476],
[1.5153],
[4.6717],
[0.7600]])
tensor([[2.6821],
[4.2126],
[3.1923],
[0.6415]], grad_fn=<MmBackward>)
tensor([[3.0468],
[3.9853],
[2.5200],
[0.3657]])
tensor([[1.9879],
[0.1151]], grad_fn=<MmBackward>)
tensor([[2.3835],
[0.1019]])
tensor([[2.0573],
[1.5737],
[1.0900],
[0.1228]], grad_fn=<MmBackward>)
tensor([[2.3835],
[1.5153],
[0.7600],
[0.1019]])
tensor([[4.0267],
[0.6093],
[3.5385],
[3.0503]], grad_fn=<MmBackward>)
tensor([[3.9853],
[0.3657],
[3.6476],
[2.5200]])
tensor([[4.2382],
[2.4021]], grad_fn=<MmBackward>)
tensor([[4.6717],
[3.0468]])
tensor([[3.2994],
[3.8279],
[0.1286],
[4.8849]], grad_fn=<MmBackward>)
tensor([[2.5200],
[3.6476],
[0.1019],
[4.6717]])
tensor([[1.4602],
[3.7197],
[2.3640],
[1.9121]], grad_fn=<MmBackward>)
tensor([[1.5153],
[3.9853],
[3.0468],
[2.3835]])
tensor([[1.1858],
[0.6599]], grad_fn=<MmBackward>)
tensor([[0.7600],
[0.3657]])
tensor([[1.1491],
[3.7226],
[1.6638],
[4.2373]], grad_fn=<MmBackward>)
tensor([[0.7600],
[3.6476],
[1.5153],
[3.9853]])
tensor([[4.4045],
[0.1024],
[2.0144],
[2.9704]], grad_fn=<MmBackward>)
tensor([[4.6717],
[0.1019],
[2.3835],
[2.5200]])
tensor([[2.5536],
[0.5956]], grad_fn=<MmBackward>)
tensor([[3.0468],
[0.3657]])
tensor([[3.1792],
[0.1114],
[4.7131],
[2.1566]], grad_fn=<MmBackward>)
tensor([[2.5200],
[0.1019],
[4.6717],
[2.3835]])
tensor([[2.4916],
[3.4475],
[1.5356],
[1.0576]], grad_fn=<MmBackward>)
tensor([[3.0468],
[3.6476],
[1.5153],
[0.7600]])
tensor([[4.2088],
[0.6227]], grad_fn=<MmBackward>)
tensor([[3.9853],
[0.3657]])
tensor([[3.0550],
[0.1008],
[0.5932],
[2.0703]], grad_fn=<MmBackward>)
tensor([[2.5200],
[0.1019],
[0.3657],
[2.3835]])
tensor([[3.3893],
[1.5051],
[1.0340],
[3.8604]], grad_fn=<MmBackward>)
tensor([[3.6476],
[1.5153],
[0.7600],
[3.9853]])
tensor([[2.5612],
[4.5348]], grad_fn=<MmBackward>)
tensor([[3.0468],
[4.6717]])
tensor([[3.8102],
[4.3393],
[1.6939],
[1.1649]], grad_fn=<MmBackward>)
tensor([[3.6476],
[3.9853],
[1.5153],
[0.7600]])
tensor([[2.9483],
[1.9937],
[0.0847],
[2.4710]], grad_fn=<MmBackward>)
tensor([[2.5200],
[2.3835],
[0.1019],
[3.0468]])
tensor([[4.5550],
[0.5913]], grad_fn=<MmBackward>)
tensor([[4.6717],
[0.3657]])
tensor([[3.1147],
[2.1077],
[0.5971],
[1.6042]], grad_fn=<MmBackward>)
tensor([[2.5200],
[2.3835],
[0.3657],
[1.5153]])
tensor([[3.8779],
[3.4032],
[4.3525],
[0.0808]], grad_fn=<MmBackward>)
tensor([[3.9853],
[3.6476],
[4.6717],
[0.1019]])
tensor([[2.7331],
[1.1501]], grad_fn=<MmBackward>)
tensor([[3.0468],
[0.7600]])
tensor([[0.6285],
[1.6993],
[3.3054],
[0.0932]], grad_fn=<MmBackward>)
tensor([[0.3657],
[1.5153],
[2.5200],
[0.1019]])
tensor([[2.4762],
[3.4392],
[3.9207],
[1.0317]], grad_fn=<MmBackward>)
tensor([[3.0468],
[3.6476],
[3.9853],
[0.7600]])
tensor([[4.7897],
[2.1732]], grad_fn=<MmBackward>)
tensor([[4.6717],
[2.3835]])
Epoch = 40
tensor([[2.6876],
[3.7299],
[0.6031],
[1.6454]], grad_fn=<MmBackward>)
tensor([[3.0468],
[3.6476],
[0.3657],
[1.5153]])
Batch = 0, Error = 0.20905181765556335
tensor([[0.0802],
[1.1340],
[4.2954],
[4.8223]], grad_fn=<MmBackward>)
tensor([[0.1019],
[0.7600],
[3.9853],
[4.6717]])
Batch = 1, Error = 0.2591085433959961
tensor([[1.9928],
[2.9573]], grad_fn=<MmBackward>)
tensor([[2.3835],
[2.5200]])
Batch = 2, Error = 0.34384119510650635
tensor([[0.5349],
[0.0630],
[3.8381],
[3.3662]], grad_fn=<MmBackward>)
tensor([[0.3657],
[0.1019],
[3.9853],
[3.6476]])
tensor([[2.0726],
[4.5773],
[2.5736],
[3.0745]], grad_fn=<MmBackward>)
tensor([[2.3835],
[4.6717],
[3.0468],
[2.5200]])
tensor([[1.0994],
[1.6113]], grad_fn=<MmBackward>)
tensor([[0.7600],
[1.5153]])
tensor([[1.5743],
[0.5693],
[0.0667],
[1.0718]], grad_fn=<MmBackward>)
tensor([[1.5153],
[0.3657],
[0.1019],
[0.7600]])
tensor([[3.0125],
[3.9980],
[2.5197],
[3.5052]], grad_fn=<MmBackward>)
tensor([[2.5200],
[3.9853],
[3.0468],
[3.6476]])
tensor([[2.0527],
[4.5446]], grad_fn=<MmBackward>)
tensor([[2.3835],
[4.6717]])
tensor([[1.6357],
[2.6805],
[3.2029],
[2.1581]], grad_fn=<MmBackward>)
tensor([[1.5153],
[3.0468],
[2.5200],
[2.3835]])
tensor([[1.0754],
[0.0641],
[3.6034],
[4.1090]], grad_fn=<MmBackward>)
tensor([[0.7600],
[0.1019],
[3.6476],
[3.9853]])
tensor([[0.5498],
[4.4925]], grad_fn=<MmBackward>)
tensor([[0.3657],
[4.6717]])
tensor([[1.0704],
[0.5637],
[1.5772],
[2.5907]], grad_fn=<MmBackward>)
tensor([[0.7600],
[0.3657],
[1.5153],
[3.0468]])
tensor([[0.0546],
[3.1698],
[3.6890],
[4.7274]], grad_fn=<MmBackward>)
tensor([[0.1019],
[2.5200],
[3.6476],
[4.6717]])
tensor([[1.9346],
[3.8285]], grad_fn=<MmBackward>)
tensor([[2.3835],
[3.9853]])
tensor([[3.5753],
[1.0592],
[2.0656],
[2.5688]], grad_fn=<MmBackward>)
tensor([[3.6476],
[0.7600],
[2.3835],
[3.0468]])
tensor([[0.0641],
[0.6021],
[4.3682],
[3.2922]], grad_fn=<MmBackward>)
tensor([[0.1019],
[0.3657],
[3.9853],
[2.5200]])
tensor([[1.4191],
[4.1832]], grad_fn=<MmBackward>)
tensor([[1.5153],
[4.6717]])
tensor([[4.6059],
[3.0869],
[1.5678],
[1.0615]], grad_fn=<MmBackward>)
tensor([[4.6717],
[2.5200],
[1.5153],
[0.7600]])
tensor([[0.0316],
[3.8041],
[0.5032],
[1.9179]], grad_fn=<MmBackward>)
tensor([[0.1019],
[3.9853],
[0.3657],
[2.3835]])
tensor([[3.5608],
[2.5558]], grad_fn=<MmBackward>)
tensor([[3.6476],
[3.0468]])
tensor([[0.0548],
[2.7166],
[4.3137],
[3.7813]], grad_fn=<MmBackward>)
tensor([[0.1019],
[3.0468],
[3.9853],
[3.6476]])
tensor([[1.5943],
[3.1355],
[2.1080],
[4.6766]], grad_fn=<MmBackward>)
tensor([[1.5153],
[2.5200],
[2.3835],
[4.6717]])
tensor([[1.0161],
[0.5303]], grad_fn=<MmBackward>)
tensor([[0.7600],
[0.3657]])
tensor([[2.9110],
[1.4736],
[3.3902],
[4.3484]], grad_fn=<MmBackward>)
tensor([[2.5200],
[1.5153],
[3.6476],
[4.6717]])
tensor([[0.5442],
[2.5579],
[4.0682],
[1.0477]], grad_fn=<MmBackward>)
tensor([[0.3657],
[3.0468],
[3.9853],
[0.7600]])
tensor([[2.0934],
[0.0396]], grad_fn=<MmBackward>)
tensor([[2.3835],
[0.1019]])
tensor([[2.6704],
[0.0467],
[3.7199],
[0.5714]], grad_fn=<MmBackward>)
tensor([[3.0468],
[0.1019],
[3.6476],
[0.3657]])
tensor([[2.1944],
[1.6582],
[4.3390],
[1.1221]], grad_fn=<MmBackward>)
tensor([[2.3835],
[1.5153],
[3.9853],
[0.7600]])
tensor([[4.5788],
[3.0646]], grad_fn=<MmBackward>)
tensor([[4.6717],
[2.5200]])
Epoch = 50
tensor([[1.4704],
[1.9514],
[4.3565],
[0.9894]], grad_fn=<MmBackward>)
tensor([[1.5153],
[2.3835],
[4.6717],
[0.7600]])
Batch = 0, Error = 0.34073713421821594
tensor([[4.2175],
[2.6504],
[0.5609],
[3.1728]], grad_fn=<MmBackward>)
tensor([[3.9853],
[3.0468],
[0.3657],
[2.5200]])
Batch = 1, Error = 0.6752656102180481
tensor([[0.0249],
[3.4094]], grad_fn=<MmBackward>)
tensor([[0.1019],
[3.6476]])
Batch = 2, Error = 0.06266370415687561
tensor([[0.0312],
[1.5305],
[2.0302],
[3.5295]], grad_fn=<MmBackward>)
tensor([[0.1019],
[1.5153],
[2.3835],
[3.6476]])
tensor([[1.0840],
[3.1686],
[4.7320],
[4.2109]], grad_fn=<MmBackward>)
tensor([[0.7600],
[2.5200],
[4.6717],
[3.9853]])
tensor([[0.4706],
[2.2866]], grad_fn=<MmBackward>)
tensor([[0.3657],
[3.0468]])
tensor([[0.0297],
[1.4998],
[2.4799],
[3.9500]], grad_fn=<MmBackward>)
tensor([[0.1019],
[1.5153],
[3.0468],
[3.9853]])
tensor([[3.6896],
[2.1270],
[0.5644],
[4.7314]], grad_fn=<MmBackward>)
tensor([[3.6476],
[2.3835],
[0.3657],
[4.6717]])
tensor([[3.1677],
[1.0843]], grad_fn=<MmBackward>)
tensor([[2.5200],
[0.7600]])
tensor([[4.3129],
[2.4064],
[1.9297],
[2.8830]], grad_fn=<MmBackward>)
tensor([[4.6717],
[3.0468],
[2.3835],
[2.5200]])
tensor([[1.1166],
[4.3313],
[0.0450],
[1.6523]], grad_fn=<MmBackward>)
tensor([[0.7600],
[3.9853],
[0.1019],
[1.5153]])
tensor([[0.5272],
[3.5143]], grad_fn=<MmBackward>)
tensor([[0.3657],
[3.6476]])
tensor([[1.0395],
[1.5449],
[4.5771],
[0.5341]], grad_fn=<MmBackward>)
tensor([[0.7600],
[1.5153],
[4.6717],
[0.3657]])
tensor([[2.0440],
[3.5611],
[3.0554],
[0.0211]], grad_fn=<MmBackward>)
tensor([[2.3835],
[3.6476],
[2.5200],
[0.1019]])
tensor([[2.4882],
[3.9689]], grad_fn=<MmBackward>)
tensor([[3.0468],
[3.9853]])
tensor([[3.1643],
[2.1202],
[2.6423],
[1.0761]], grad_fn=<MmBackward>)
tensor([[2.5200],
[2.3835],
[3.0468],
[0.7600]])
tensor([[0.0262],
[1.5507],
[3.5835],
[0.5344]], grad_fn=<MmBackward>)
tensor([[0.1019],
[1.5153],
[3.6476],
[0.3657]])
tensor([[4.6138],
[4.1039]], grad_fn=<MmBackward>)
tensor([[4.6717],
[3.9853]])
tensor([[0.5294],
[1.5408],
[2.0465],
[0.0237]], grad_fn=<MmBackward>)
tensor([[0.3657],
[1.5153],
[2.3835],
[0.1019]])
tensor([[2.6107],
[4.6768],
[1.0612],
[4.1603]], grad_fn=<MmBackward>)
tensor([[3.0468],
[4.6717],
[0.7600],
[3.9853]])
tensor([[3.6519],
[3.1341]], grad_fn=<MmBackward>)
tensor([[3.6476],
[2.5200]])
tensor([[3.8677],
[3.3861],
[1.4597],
[0.4965]], grad_fn=<MmBackward>)
tensor([[3.9853],
[3.6476],
[1.5153],
[0.3657]])
tensor([[4.6017],
[2.0569],
[2.5658],
[1.0389]], grad_fn=<MmBackward>)
tensor([[4.6717],
[2.3835],
[3.0468],
[0.7600]])
tensor([[3.3081],
[0.0330]], grad_fn=<MmBackward>)
tensor([[2.5200],
[0.1019]])
tensor([[2.5174],
[2.0176],
[3.5169],
[0.0186]], grad_fn=<MmBackward>)
tensor([[3.0468],
[2.3835],
[3.6476],
[0.1019]])
tensor([[4.9795],
[0.5895],
[1.6870],
[4.4307]], grad_fn=<MmBackward>)
tensor([[4.6717],
[0.3657],
[1.5153],
[3.9853]])
tensor([[2.8967],
[0.9774]], grad_fn=<MmBackward>)
tensor([[2.5200],
[0.7600]])
tensor([[3.6343],
[3.1807],
[2.2736],
[1.8201]], grad_fn=<MmBackward>)
tensor([[3.9853],
[3.6476],
[3.0468],
[2.3835]])
tensor([[3.4836],
[1.7663],
[0.0490],
[1.1939]], grad_fn=<MmBackward>)
tensor([[2.5200],
[1.5153],
[0.1019],
[0.7600]])
tensor([[4.5196],
[0.5174]], grad_fn=<MmBackward>)
tensor([[4.6717],
[0.3657]])
Epoch = 60
tensor([[0.0171],
[3.6021],
[2.0657],
[0.5293]], grad_fn=<MmBackward>)
tensor([[0.1019],
[3.6476],
[2.3835],
[0.3657]])
Batch = 0, Error = 0.1369948536157608
tensor([[1.0749],
[4.2312],
[1.6009],
[4.7572]], grad_fn=<MmBackward>)
tensor([[0.7600],
[3.9853],
[1.5153],
[4.6717]])
Batch = 1, Error = 0.17421618103981018
tensor([[2.9526],
[2.4618]], grad_fn=<MmBackward>)
tensor([[2.5200],
[3.0468]])
Batch = 2, Error = 0.5293147563934326
tensor([[2.9749],
[2.4809],
[0.0112],
[0.9991]], grad_fn=<MmBackward>)
tensor([[2.5200],
[3.0468],
[0.1019],
[0.7600]])
tensor([[1.9715],
[0.5007],
[4.4228],
[1.4812]], grad_fn=<MmBackward>)
tensor([[2.3835],
[0.3657],
[4.6717],
[1.5153]])
tensor([[3.7166],
[4.2444]], grad_fn=<MmBackward>)
tensor([[3.6476],
[3.9853]])
tensor([[3.5357],
[1.5239],
[4.0386],
[3.0327]], grad_fn=<MmBackward>)
tensor([[3.6476],
[1.5153],
[3.9853],
[2.5200]])
tensor([[0.4820],
[4.2915],
[0.0058],
[2.3868]], grad_fn=<MmBackward>)
tensor([[0.3657],
[4.6717],
[0.1019],
[3.0468]])
tensor([[1.1074],
[2.1886]], grad_fn=<MmBackward>)
tensor([[0.7600],
[2.3835]])
tensor([[2.7302],
[0.0232],
[1.1060],
[2.1888]], grad_fn=<MmBackward>)
tensor([[3.0468],
[0.1019],
[0.7600],
[2.3835]])
tensor([[3.9318],
[1.7011],
[3.3742],
[0.5858]], grad_fn=<MmBackward>)
tensor([[3.6476],
[1.5153],
[2.5200],
[0.3657]])
tensor([[3.8432],
[4.3239]], grad_fn=<MmBackward>)
tensor([[3.9853],
[4.6717]])
tensor([[4.7081],
[2.6187],
[0.0070],
[4.1858]], grad_fn=<MmBackward>)
tensor([[4.6717],
[3.0468],
[0.1019],
[3.9853]])
tensor([[3.1591],
[2.1103],
[0.5371],
[1.5859]], grad_fn=<MmBackward>)
tensor([[2.5200],
[2.3835],
[0.3657],
[1.5153]])
tensor([[0.9884],
[3.4581]], grad_fn=<MmBackward>)
tensor([[0.7600],
[3.6476]])
tensor([[ 4.5215],
[ 2.5118],
[ 1.0046],
[-0.0002]], grad_fn=<MmBackward>)
tensor([[4.6717],
[3.0468],
[0.7600],
[0.1019]])
tensor([[4.3058],
[0.5475],
[3.7689],
[2.1582]], grad_fn=<MmBackward>)
tensor([[3.9853],
[0.3657],
[3.6476],
[2.3835]])
tensor([[1.5346],
[3.0665]], grad_fn=<MmBackward>)
tensor([[1.5153],
[2.5200]])
tensor([[ 1.4257],
[ 1.9038],
[ 3.8162],
[-0.0086]], grad_fn=<MmBackward>)
tensor([[1.5153],
[2.3835],
[3.9853],
[0.1019]])
tensor([[3.5967],
[3.0841],
[2.5714],
[4.6219]], grad_fn=<MmBackward>)
tensor([[3.6476],
[2.5200],
[3.0468],
[4.6717]])
tensor([[0.5192],
[1.0299]], grad_fn=<MmBackward>)
tensor([[0.3657],
[0.7600]])
tensor([[1.5118],
[4.5351],
[4.0312],
[3.5273]], grad_fn=<MmBackward>)
tensor([[1.5153],
[4.6717],
[3.9853],
[3.6476]])
tensor([[3.1281],
[1.0456],
[0.5250],
[2.0868]], grad_fn=<MmBackward>)
tensor([[2.5200],
[0.7600],
[0.3657],
[2.3835]])
tensor([[ 2.4367],
[-0.0107]], grad_fn=<MmBackward>)
tensor([[3.0468],
[0.1019]])
tensor([[2.5999],
[2.0806],
[4.1575],
[1.0422]], grad_fn=<MmBackward>)
tensor([[3.0468],
[2.3835],
[3.9853],
[0.7600]])
tensor([[0.0096],
[1.6113],
[3.2130],
[3.7469]], grad_fn=<MmBackward>)
tensor([[0.1019],
[1.5153],
[2.5200],
[3.6476]])
tensor([[4.3477],
[0.4775]], grad_fn=<MmBackward>)
tensor([[4.6717],
[0.3657]])
tensor([[ 0.5091],
[ 2.5535],
[-0.0020],
[ 1.5313]], grad_fn=<MmBackward>)
tensor([[0.3657],
[3.0468],
[0.1019],
[1.5153]])
tensor([[2.1399],
[3.7397],
[3.2065],
[4.2730]], grad_fn=<MmBackward>)
tensor([[2.3835],
[3.6476],
[2.5200],
[3.9853]])
tensor([[0.9381],
[4.2555]], grad_fn=<MmBackward>)
tensor([[0.7600],
[4.6717]])
Epoch = 70
tensor([[4.0507],
[1.5159],
[1.0089],
[4.5577]], grad_fn=<MmBackward>)
tensor([[3.9853],
[1.5153],
[0.7600],
[4.6717]])
Batch = 0, Error = 0.07924719899892807
tensor([[ 2.0190],
[ 2.5260],
[-0.0090],
[ 0.4980]], grad_fn=<MmBackward>)
tensor([[2.3835],
[3.0468],
[0.1019],
[0.3657]])
Batch = 1, Error = 0.4338834285736084
tensor([[3.2801],
[3.8254]], grad_fn=<MmBackward>)
tensor([[2.5200],
[3.6476]])
Batch = 2, Error = 0.6094035506248474
tensor([[3.8992],
[1.9444],
[1.4557],
[2.4331]], grad_fn=<MmBackward>)
tensor([[3.9853],
[2.3835],
[1.5153],
[3.0468]])
tensor([[4.9113],
[1.1019],
[3.8229],
[3.2787]], grad_fn=<MmBackward>)
tensor([[4.6717],
[0.7600],
[3.6476],
[2.5200]])
tensor([[ 0.4434],
[-0.0168]], grad_fn=<MmBackward>)
tensor([[0.3657],
[0.1019]])
tensor([[ 0.9029],
[-0.0160],
[ 1.3623],
[ 1.8217]], grad_fn=<MmBackward>)
tensor([[0.7600],
[0.1019],
[1.5153],
[2.3835]])
tensor([[2.8959],
[0.4808],
[3.8619],
[2.4129]], grad_fn=<MmBackward>)
tensor([[2.5200],
[0.3657],
[3.9853],
[3.0468]])
tensor([[3.5061],
[4.5070]], grad_fn=<MmBackward>)
tensor([[3.6476],
[4.6717]])
tensor([[2.1074],
[2.6319],
[4.7301],
[3.6810]], grad_fn=<MmBackward>)
tensor([[2.3835],
[3.0468],
[4.6717],
[3.6476]])
tensor([[0.5694],
[0.0212],
[1.6656],
[1.1175]], grad_fn=<MmBackward>)
tensor([[0.3657],
[0.1019],
[1.5153],
[0.7600]])
tensor([[3.2173],
[4.2868]], grad_fn=<MmBackward>)
tensor([[2.5200],
[3.9853]])
tensor([[1.8706],
[1.4001],
[3.2820],
[0.9296]], grad_fn=<MmBackward>)
tensor([[2.3835],
[1.5153],
[3.6476],
[0.7600]])
tensor([[2.5826],
[4.1290],
[3.0981],
[0.0051]], grad_fn=<MmBackward>)
tensor([[3.0468],
[3.9853],
[2.5200],
[0.1019]])
tensor([[0.4950],
[4.4398]], grad_fn=<MmBackward>)
tensor([[0.3657],
[4.6717]])
tensor([[2.5649],
[3.0771],
[1.0283],
[1.5405]], grad_fn=<MmBackward>)
tensor([[3.0468],
[2.5200],
[0.7600],
[1.5153]])
tensor([[ 3.4764],
[-0.0034],
[ 1.9851],
[ 0.4937]], grad_fn=<MmBackward>)
tensor([[3.6476],
[0.1019],
[2.3835],
[0.3657]])
tensor([[4.7153],
[4.1922]], grad_fn=<MmBackward>)
tensor([[4.6717],
[3.9853]])
tensor([[2.0150],
[3.5244],
[3.0213],
[0.5056]], grad_fn=<MmBackward>)
tensor([[2.3835],
[3.6476],
[2.5200],
[0.3657]])
tensor([[ 4.4565],
[ 2.4756],
[-0.0005],
[ 1.4852]], grad_fn=<MmBackward>)
tensor([[4.6717],
[3.0468],
[0.1019],
[1.5153]])
tensor([[1.1035],
[4.3605]], grad_fn=<MmBackward>)
tensor([[0.7600],
[3.9853]])
tensor([[0.5104],
[4.0585],
[4.5654],
[3.5516]], grad_fn=<MmBackward>)
tensor([[0.3657],
[3.9853],
[4.6717],
[3.6476]])
tensor([[2.5814],
[1.0345],
[3.0970],
[1.5501]], grad_fn=<MmBackward>)
tensor([[3.0468],
[0.7600],
[2.5200],
[1.5153]])
tensor([[-0.0052],
[ 1.9876]], grad_fn=<MmBackward>)
tensor([[0.1019],
[2.3835]])
tensor([[4.6276],
[3.6003],
[3.0867],
[0.0049]], grad_fn=<MmBackward>)
tensor([[4.6717],
[3.6476],
[2.5200],
[0.1019]])
tensor([[0.4849],
[1.4601],
[2.4353],
[3.8980]], grad_fn=<MmBackward>)
tensor([[0.3657],
[1.5153],
[3.0468],
[3.9853]])
tensor([[1.0593],
[2.1086]], grad_fn=<MmBackward>)
tensor([[0.7600],
[2.3835]])
tensor([[3.7163],
[2.6572],
[4.7754],
[1.0686]], grad_fn=<MmBackward>)
tensor([[3.6476],
[3.0468],
[4.6717],
[0.7600]])
tensor([[0.5364],
[2.1226],
[3.1801],
[0.0077]], grad_fn=<MmBackward>)
tensor([[0.3657],
[2.3835],
[2.5200],
[0.1019]])
tensor([[1.4941],
[3.9873]], grad_fn=<MmBackward>)
tensor([[1.5153],
[3.9853]])
Epoch = 80
tensor([[ 2.9932],
[-0.0015],
[ 3.4923],
[ 4.4905]], grad_fn=<MmBackward>)
tensor([[2.5200],
[0.1019],
[3.6476],
[4.6717]])
Batch = 0, Error = 0.29155468940734863
tensor([[1.4917],
[1.9896],
[3.9813],
[0.4958]], grad_fn=<MmBackward>)
tensor([[1.5153],
[2.3835],
[3.9853],
[0.3657]])
Batch = 1, Error = 0.1726582795381546
tensor([[2.5688],
[1.0298]], grad_fn=<MmBackward>)
tensor([[3.0468],
[0.7600]])
Batch = 2, Error = 0.3012203872203827
tensor([[2.6632],
[1.6011],
[3.1942],
[3.7253]], grad_fn=<MmBackward>)
tensor([[3.0468],
[1.5153],
[2.5200],
[3.6476]])
tensor([[1.0038],
[0.5013],
[2.0088],
[4.0189]], grad_fn=<MmBackward>)
tensor([[0.7600],
[0.3657],
[2.3835],
[3.9853]])
tensor([[ 4.5739],
[-0.0020]], grad_fn=<MmBackward>)
tensor([[4.6717],
[0.1019]])
tensor([[2.5871],
[3.1041],
[2.0701],
[0.5191]], grad_fn=<MmBackward>)
tensor([[3.0468],
[2.5200],
[2.3835],
[0.3657]])
tensor([[1.0347],
[1.5507],
[0.0028],
[4.6466]], grad_fn=<MmBackward>)
tensor([[0.7600],
[1.5153],
[0.1019],
[4.6717]])
tensor([[4.0934],
[3.5816]], grad_fn=<MmBackward>)
tensor([[3.9853],
[3.6476]])
tensor([[2.0297],
[0.5061],
[4.5690],
[1.0140]], grad_fn=<MmBackward>)
tensor([[2.3835],
[0.3657],
[4.6717],
[0.7600]])
tensor([[-0.0005],
[ 3.1456],
[ 2.6212],
[ 4.1943]], grad_fn=<MmBackward>)
tensor([[0.1019],
[2.5200],
[3.0468],
[3.9853]])
tensor([[1.4700],
[3.4388]], grad_fn=<MmBackward>)
tensor([[1.5153],
[3.6476]])
tensor([[4.5685],
[2.0295],
[3.5529],
[1.5218]], grad_fn=<MmBackward>)
tensor([[4.6717],
[2.3835],
[3.6476],
[1.5153]])
tensor([[0.0093],
[4.3047],
[2.6939],
[3.2308]], grad_fn=<MmBackward>)
tensor([[0.1019],
[3.9853],
[3.0468],
[2.5200]])
tensor([[0.4853],
[0.9729]], grad_fn=<MmBackward>)
tensor([[0.3657],
[0.7600]])
tensor([[-0.0090],
[ 3.3673],
[ 4.3319],
[ 2.4026]], grad_fn=<MmBackward>)
tensor([[0.1019],
[3.6476],
[4.6717],
[3.0468]])
tensor([[3.3945],
[2.2691],
[1.7065],
[1.1438]], grad_fn=<MmBackward>)
tensor([[2.5200],
[2.3835],
[1.5153],
[0.7600]])
tensor([[4.0150],
[0.4947]], grad_fn=<MmBackward>)
tensor([[3.9853],
[0.3657]])
tensor([[1.9859],
[4.4826],
[2.4853],
[2.9846]], grad_fn=<MmBackward>)
tensor([[2.3835],
[4.6717],
[3.0468],
[2.5200]])
tensor([[1.0655],
[1.5971],
[3.7236],
[4.2553]], grad_fn=<MmBackward>)
tensor([[0.7600],
[1.5153],
[3.6476],
[3.9853]])
tensor([[-0.0124],
[ 0.4846]], grad_fn=<MmBackward>)
tensor([[0.1019],
[0.3657]])
tensor([[2.4669],
[3.9545],
[1.9710],
[0.9793]], grad_fn=<MmBackward>)
tensor([[3.0468],
[3.9853],
[2.3835],
[0.7600]])
tensor([[0.0036],
[0.5419],
[3.2337],
[1.6187]], grad_fn=<MmBackward>)
tensor([[0.1019],
[0.3657],
[2.5200],
[1.5153]])
tensor([[3.4288],
[4.4125]], grad_fn=<MmBackward>)
tensor([[3.6476],
[4.6717]])
tensor([[ 4.7611],
[-0.0047],
[ 3.1725],
[ 1.0543]], grad_fn=<MmBackward>)
tensor([[4.6717],
[0.1019],
[2.5200],
[0.7600]])
tensor([[3.7990],
[2.3657],
[1.8879],
[1.4101]], grad_fn=<MmBackward>)
tensor([[3.9853],
[3.0468],
[2.3835],
[1.5153]])
tensor([[0.5540],
[3.8415]], grad_fn=<MmBackward>)
tensor([[0.3657],
[3.6476]])
tensor([[2.6627],
[1.0641],
[0.5312],
[1.5970]], grad_fn=<MmBackward>)
tensor([[3.0468],
[0.7600],
[0.3657],
[1.5153]])
tensor([[ 4.3283],
[ 2.1616],
[-0.0050],
[ 3.7866]], grad_fn=<MmBackward>)
tensor([[3.9853],
[2.3835],
[0.1019],
[3.6476]])
tensor([[4.6187],
[3.0764]], grad_fn=<MmBackward>)
tensor([[4.6717],
[2.5200]])
Epoch = 90
tensor([[ 2.4129],
[ 2.8990],
[ 0.9543],
[-0.0181]], grad_fn=<MmBackward>)
tensor([[3.0468],
[2.5200],
[0.7600],
[0.1019]])
Batch = 0, Error = 0.5977063179016113
tensor([[1.9500],
[0.4767],
[4.4057],
[1.4589]], grad_fn=<MmBackward>)
tensor([[2.3835],
[0.3657],
[4.6717],
[1.5153]])
Batch = 1, Error = 0.27414506673812866
tensor([[4.2539],
[3.7220]], grad_fn=<MmBackward>)
tensor([[3.9853],
[3.6476]])
Batch = 2, Error = 0.07766690850257874
tensor([[4.0389],
[1.0034],
[3.5330],
[0.4975]], grad_fn=<MmBackward>)
tensor([[3.9853],
[0.7600],
[3.6476],
[0.3657]])
tensor([[ 4.5171],
[ 2.5029],
[ 3.0065],
[-0.0147]], grad_fn=<MmBackward>)
tensor([[4.6717],
[3.0468],
[2.5200],
[0.1019]])
tensor([[2.0525],
[1.5373]], grad_fn=<MmBackward>)
tensor([[2.3835],
[1.5153]])
tensor([[1.5803],
[4.2174],
[2.1077],
[3.1625]], grad_fn=<MmBackward>)
tensor([[1.5153],
[3.9853],
[2.3835],
[2.5200]])
tensor([[-0.0152],
[ 0.9459],
[ 2.3877],
[ 4.3101]], grad_fn=<MmBackward>)
tensor([[0.1019],
[0.7600],
[3.0468],
[4.6717]])
tensor([[3.7895],
[0.5446]], grad_fn=<MmBackward>)
tensor([[3.6476],
[0.3657]])
tensor([[4.7619],
[0.5268],
[4.2325],
[1.5856]], grad_fn=<MmBackward>)
tensor([[4.6717],
[0.3657],
[3.9853],
[1.5153]])
tensor([[ 1.9803],
[ 2.9775],
[ 0.9832],
[-0.0140]], grad_fn=<MmBackward>)
tensor([[2.3835],
[2.5200],
[0.7600],
[0.1019]])
tensor([[2.3987],
[3.3651]], grad_fn=<MmBackward>)
tensor([[3.0468],
[3.6476]])
tensor([[0.5354],
[4.8079],
[1.0695],
[0.0014]], grad_fn=<MmBackward>)
tensor([[0.3657],
[4.6717],
[0.7600],
[0.1019]])
tensor([[4.1065],
[2.0488],
[1.5344],
[3.0776]], grad_fn=<MmBackward>)
tensor([[3.9853],
[2.3835],
[1.5153],
[2.5200]])
tensor([[2.4081],
[3.3778]], grad_fn=<MmBackward>)
tensor([[3.0468],
[3.6476]])
tensor([[1.6052],
[2.1396],
[4.2771],
[3.2084]], grad_fn=<MmBackward>)
tensor([[1.5153],
[2.3835],
[3.9853],
[2.5200]])
tensor([[ 0.4637],
[ 4.2898],
[-0.0145],
[ 2.3768]], grad_fn=<MmBackward>)
tensor([[0.3657],
[4.6717],
[0.1019],
[3.0468]])
tensor([[3.8112],
[1.0938]], grad_fn=<MmBackward>)
tensor([[3.6476],
[0.7600]])
tensor([[3.6775],
[2.1001],
[2.6259],
[1.0485]], grad_fn=<MmBackward>)
tensor([[3.6476],
[2.3835],
[3.0468],
[0.7600]])
tensor([[0.5543],
[1.6537],
[0.0046],
[3.3028]], grad_fn=<MmBackward>)
tensor([[0.3657],
[1.5153],
[0.1019],
[2.5200]])
tensor([[4.4666],
[3.9686]], grad_fn=<MmBackward>)
tensor([[4.6717],
[3.9853]])
tensor([[0.5062],
[3.0928],
[1.5408],
[2.0581]], grad_fn=<MmBackward>)
tensor([[0.3657],
[2.5200],
[1.5153],
[2.3835]])
tensor([[0.9694],
[4.4301],
[3.9357],
[3.4413]], grad_fn=<MmBackward>)
tensor([[0.7600],
[4.6717],
[3.9853],
[3.6476]])
tensor([[ 2.6336],
[-0.0137]], grad_fn=<MmBackward>)
tensor([[3.0468],
[0.1019]])
tensor([[ 0.5465],
[ 2.7449],
[ 2.1953],
[-0.0031]], grad_fn=<MmBackward>)
tensor([[0.3657],
[3.0468],
[2.3835],
[0.1019]])
tensor([[1.7148],
[5.1342],
[3.9944],
[3.4245]], grad_fn=<MmBackward>)
tensor([[1.5153],
[4.6717],
[3.6476],
[2.5200]])
tensor([[3.5423],
[0.8608]], grad_fn=<MmBackward>)
tensor([[3.9853],
[0.7600]])
tensor([[0.9328],
[2.3713],
[0.4533],
[2.8508]], grad_fn=<MmBackward>)
tensor([[0.7600],
[3.0468],
[0.3657],
[2.5200]])
tensor([[ 1.4420],
[ 1.9309],
[-0.0245],
[ 3.3975]], grad_fn=<MmBackward>)
tensor([[1.5153],
[2.3835],
[0.1019],
[3.6476]])
tensor([[4.1992],
[4.7250]], grad_fn=<MmBackward>)
tensor([[3.9853],
[4.6717]])
Lets see how well the model has learnt the data¶
In [14]:
#from torch.autograd.variable import Variable
x_for_plotting = np.linspace(0, 2*np.pi, 1000)
design_matrix = torch.tensor(np.vstack([np.ones(x_for_plotting.shape), x_for_plotting]).T, dtype=torch.float32)
#print('Design matrix:\n', design_matrix)
print('Design matrix shape:', design_matrix.shape)
y_for_plotting = model.forward(design_matrix)
print('y_for_plotting shape:', y_for_plotting.shape)
Design matrix shape: torch.Size([1000, 2]) y_for_plotting shape: torch.Size([1000, 1])
In [15]:
plt.figure()
plt.plot(x,y,'o')
plt.plot(x_for_plotting, y_for_plotting.data.numpy(), 'r-')
plt.xlabel('x')
plt.ylabel('y')
plt.title('2D data (#data = %d)' % N)
plt.show()
