Tensorflow, how to chain GRU layers - neural-network

Right now I am trying to chain multiple GRU recurrent layers to each other in tensorflow. I am getting the following error.
ValueError: Variable GRUCell/Gates/Linear/Matrix already exists, disallowed. Did you mean to set reuse=True in VarScope? Originally defined at:
File "/home/chase/workspace/SentenceEncoder/sent_enc.py", line 42, in <module>
output, states[i] = grus[i](output, states[i])
Here is my code.
x = tf.placeholder(tf.float32, (batch_size, time_steps, vlen), 'x')
y_exp = tf.placeholder(tf.float32, (batch_size, time_steps, vlen), 'y_exp')
with tf.name_scope('encoder'):
gru_sizes = (128, 256, 512)
grus = [tf.nn.rnn_cell.GRUCell(sz) for sz in gru_sizes]
states = [tf.zeros((batch_size, g.state_size)) for g in grus]
for t in range(time_steps):
output = tf.reshape(x[:, t, :], (batch_size, vlen))
for i in range(len(grus)):
output, states[i] = grus[i](output, states[i])
I am aware that tensorflow provides a MultiRNNCell for doing this but I kind of wanted to figure it out for myself.

I managed to fix it. I needed to add a different variable scope for each of the layers. I also needed to reuse the variables after the first time step.
x = tf.placeholder(tf.float32, (batch_size, time_steps, vlen), 'x')
y_exp = tf.placeholder(tf.float32, (batch_size, time_steps, vlen), 'y_exp')
with tf.name_scope('encoder'):
gru_sizes = (128, 256, 512)
grus = [tf.nn.rnn_cell.GRUCell(sz) for sz in gru_sizes]
states = [tf.zeros((batch_size, g.state_size)) for g in grus]
for t in range(time_steps):
output = tf.reshape(x[:, t, :], (batch_size, vlen))
for i in range(len(grus)):
with tf.variable_scope('gru_' + str(i), reuse = t > 0):
output, states[i] = grus[i](output, states[i])

Related

Given groups=1, weight of size [10, 1, 5, 5], expected input[2, 3, 28, 28] to have 1 channels, but got 3 channels instead

I am trying to run CNN with train MNIST, but test on my own written digits. To do that I wrote the following code but I getting an error in title of this questions:
I am trying to run CNN with train MNIST, but test on my own written digits. To do that I wrote the following code but I getting an error in title of this questions:
batch_size = 64
train_dataset = datasets.MNIST(root='./data/',
train=True,
transform=transforms.ToTensor(),
download=True)
test_dataset = ImageFolder('my_digit_images/', transform=transforms.ToTensor())
train_loader = torch.utils.data.DataLoader(dataset=train_dataset,
batch_size=batch_size,
shuffle=True)
test_loader = torch.utils.data.DataLoader(dataset=test_dataset,
batch_size=batch_size,
shuffle=False)
class Net(nn.Module):
def __init__(self):
super(Net, self).__init__()
self.conv1 = nn.Conv2d(1, 10, kernel_size=5)
#print(self.conv1.weight.shape)
self.conv2 = nn.Conv2d(10, 20, kernel_size=5)
self.conv3 = nn.Conv2d(20, 20, kernel_size=3)
#print(self.conv2.weight.shape)
self.mp = nn.MaxPool2d(2)
self.fc = nn.Linear(320, 10)
def forward(self, x):
in_size = x.size(0)
x = F.relu(self.conv1(x))
#print(x.shape)
x = F.relu(self.mp(self.conv2(x)))
x = F.relu(self.mp(self.conv3(x)))
#print("2.", x.shape)
# x = F.relu(self.mp(self.conv3(x)))
x = x.view(in_size, -1) # flatten the tensor
#print("3.", x.shape)
x = self.fc(x)
return F.log_softmax(x)
model = Net()
optimizer = optim.SGD(model.parameters(), lr=0.01, momentum=0.5)
def train(epoch):
model.train()
for batch_idx, (data, target) in enumerate(train_loader):
data, target = Variable(data), Variable(target)
optimizer.zero_grad()
output = model(data)
loss = F.nll_loss(output, target)
loss.backward()
optimizer.step()
if batch_idx % 10 == 0:
print('Train Epoch: {} [{}/{} ({:.0f}%)]\tLoss: {:.6f}'.format(
epoch, batch_idx * len(data), len(train_loader.dataset),
100. * batch_idx / len(train_loader), loss.item()))
def test():
model.eval()
test_loss = 0
correct = 0
for data, target in test_loader:
data, target = Variable(data, volatile=True), Variable(target)
output = model(data)
test_loss += F.nll_loss(output, target, size_average=False).data
pred = output.data.max(1, keepdim=True)[1]
correct += pred.eq(target.data.view_as(pred)).cpu().sum()
test_loss /= len(test_loader.dataset)
print('\nTest set: Average loss: {:.4f}, Accuracy: {}/{} ({:.0f}%)\n'.format(
test_loss, correct, len(test_loader.dataset),
100. * correct / len(test_loader.dataset)))
MNIST dataset contains black and white 1-channel images, while yours are 3-channeled RGB probably. Either recode your images or preprocess it like
img = img[:,0:1,:,:]
You can do it with custom transform, adding it after transforms.ToTensor()
The images in training and testing should follow the same distribution. Since MNIST data is by default in Grayscale and it is expected that you didn't change the channels, then the model expects the same number of channels in testing.
The following code is an example of how it's done using a transformation.
Following the order defined below, it
Converts the image to a single channel (Grayscale)
Resize the image to the size of the default MNIST data
Convert the image to a tensor
Normalize the tensor to have same mean and std as that of during training(assuming that you used the same values).
test_dataset = ImageFolder('my_digit_images/', transform=transforms.Compose([transforms.Grayscale(num_output_channels=1),
transforms.Resize((28, 28)),
transforms.ToTensor(),
transforms.Normalize((0.1307,), (0.3081,))]))

I want to use Numpy to simulate the inference process of a quantized MobileNet V2 network, but the outcome is different with pytorch realized one

Python version: 3.8
Pytorch version: 1.9.0+cpu
Platform: Anaconda Spyder5.0
To reproduce this problem, just copy every code below to a single file.
The ILSVRC2012_val_00000293.jpg file used in this code is shown below, you also need to download it and then change its destination in the code.
Some background of this problem:
I am now working on a project that aims to develop a hardware accelerator to complete the inference process of the MobileNet V2 network. I used pretrained quantized Pytorch model to simulate the outcome, and the result comes out very well.
In order to use hardware to complete this task, I wish to know every inputs and outputs as well as intermidiate variables during runing this piece of pytorch code. I used a package named torchextractor to fetch the outcomes of first layer, which in this case, is a 3*3 convolution layer.
import numpy as np
import torchvision
import torch
from torchvision import transforms, datasets
from PIL import Image
from torchvision import transforms
import torchextractor as tx
import math
#########################################################################################
##### Processing of input image
#########################################################################################
normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
test_transform = transforms.Compose([
transforms.Resize(256),
transforms.CenterCrop(224),
transforms.ToTensor(),
normalize,])
preprocess = transforms.Compose([
transforms.Resize(256),
transforms.CenterCrop(224),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
])
#image file destination
filename = "D:\Project_UM\MobileNet_VC709\MobileNet_pytorch\ILSVRC2012_val_00000293.jpg"
input_image = Image.open(filename)
input_tensor = preprocess(input_image)
input_batch = input_tensor.unsqueeze(0)
#########################################################################################
#########################################################################################
#########################################################################################
#----First verify that the torchextractor class should not influent the inference outcome
# ofmp of layer1 before putting into torchextractor
a,b,c = quantize_tensor(input_batch)# to quantize the input tensor and return an int8 tensor, scale and zero point
input_qa = torch.quantize_per_tensor(torch.tensor(input_batch.clone().detach()), b, c, torch.quint8)# Using quantize_per_tensor method of torch
# Load a quantized mobilenet_v2 model
model_quantized = torchvision.models.quantization.mobilenet_v2(pretrained=True, quantize=True)
model_quantized.eval()
with torch.no_grad():
output = model_quantized.features[0][0](input_qa)# Ofmp of layer1, datatype : quantized_tensor
# print("FM of layer1 before tx_extractor:\n",output.int_repr())# Ofmp of layer1, datatype : int8 tensor
output1_clone = output.int_repr().detach().numpy()# Clone ofmp of layer1, datatype : ndarray
#########################################################################################
#########################################################################################
#########################################################################################
# ofmp of layer1 after adding torchextractor
model_quantized_ex = tx.Extractor(model_quantized, ["features.0.0"])#Capture of the module inside first layer
model_output, features = model_quantized_ex(input_batch)# Forward propagation
# feature_shapes = {name: f.shape for name, f in features.items()}
# print(features['features.0.0']) # Ofmp of layer1, datatype : quantized_tensor
out1_clone = features['features.0.0'].int_repr().numpy() # Clone ofmp of layer1, datatype : ndarray
if(out1_clone.all() == output1_clone.all()):
print('Model with torchextractor attached output the same value as the original model')
else:
print('Torchextractor method influence the outcome')
Here I define a numpy quantization scheme based on the quantization scheme proposed by
Quantization and Training of Neural Networks for Efficient
Integer-Arithmetic-Only Inference
# Convert a normal regular tensor to a quantized tensor with scale and zero_point
def quantize_tensor(x, num_bits=8):# to quantize the input tensor and return an int8 tensor, scale and zero point
qmin = 0.
qmax = 2.**num_bits - 1.
min_val, max_val = x.min(), x.max()
scale = (max_val - min_val) / (qmax - qmin)
initial_zero_point = qmin - min_val / scale
zero_point = 0
if initial_zero_point < qmin:
zero_point = qmin
elif initial_zero_point > qmax:
zero_point = qmax
else:
zero_point = initial_zero_point
# print(zero_point)
zero_point = int(zero_point)
q_x = zero_point + x / scale
q_x.clamp_(qmin, qmax).round_()
q_x = q_x.round().byte()
return q_x, scale, zero_point
#%%
# #############################################################################################
# --------- Simulate the inference process of layer0: conv33 using numpy
# #############################################################################################
# get the input_batch quantized buffer data
input_scale = b.item()
input_zero = c
input_quantized = a[0].detach().numpy()
# get the layer0 output scale and zero_point
output_scale = model_quantized.features[0][0].state_dict()['scale'].item()
output_zero = model_quantized.features[0][0].state_dict()['zero_point'].item()
# get the quantized weight with scale and zero_point
weight_scale = model_quantized.features[0][0].state_dict()["weight"].q_scale()
weight_zero = model_quantized.features[0][0].state_dict()["weight"].q_zero_point()
weight_quantized = model_quantized.features[0][0].state_dict()["weight"].int_repr().numpy()
# print(weight_quantized)
# print(weight_quantized.shape)
# bias_quantized,bias_scale,bias_zero= quantize_tensor(model_quantized.features[0][0].state_dict()["bias"])# to quantize the input tensor and return an int8 tensor, scale and zero point
# print(bias_quantized.shape)
bias = model_quantized.features[0][0].state_dict()["bias"].detach().numpy()
# print(input_quantized)
print(type(input_scale))
print(type(output_scale))
print(type(weight_scale))
Then I write a quantized 2D convolution using numpy, hope to figure out every details in pytorch data flow during the inference.
#%% numpy simulated layer0 convolution function define
def conv_cal(input_quantized, weight_quantized, kernel_size, stride, out_i, out_j, out_k):
weight = weight_quantized[out_i]
input = np.zeros((input_quantized.shape[0], kernel_size, kernel_size))
for i in range(weight.shape[0]):
for j in range(weight.shape[1]):
for k in range(weight.shape[2]):
input[i][j][k] = input_quantized[i][stride*out_j+j][stride*out_k+k]
# print(np.dot(weight,input))
# print(input,"\n")
# print(weight)
return np.multiply(weight,input).sum()
def QuantizedConv2D(input_scale, input_zero, input_quantized, output_scale, output_zero, weight_scale, weight_zero, weight_quantized, bias, kernel_size, stride, padding, ofm_size):
output = np.zeros((weight_quantized.shape[0],ofm_size,ofm_size))
input_quantized_padding = np.full((input_quantized.shape[0],input_quantized.shape[1]+2*padding,input_quantized.shape[2]+2*padding),0)
zero_temp = np.full(input_quantized.shape,input_zero)
input_quantized = input_quantized - zero_temp
for i in range(input_quantized.shape[0]):
for j in range(padding,padding + input_quantized.shape[1]):
for k in range(padding,padding + input_quantized.shape[2]):
input_quantized_padding[i][j][k] = input_quantized[i][j-padding][k-padding]
zero_temp = np.full(weight_quantized.shape, weight_zero)
weight_quantized = weight_quantized - zero_temp
for i in range(output.shape[0]):
for j in range(output.shape[1]):
for k in range(output.shape[2]):
# output[i][j][k] = (weight_scale*input_scale)*conv_cal(input_quantized_padding, weight_quantized, kernel_size, stride, i, j, k) + bias[i] #floating_output
output[i][j][k] = weight_scale*input_scale/output_scale*conv_cal(input_quantized_padding, weight_quantized, kernel_size, stride, i, j, k) + bias[i]/output_scale + output_zero
output[i][j][k] = round(output[i][j][k])
# int_output
return output
Here I input the same image, weight, and bias together with their zero_point and scale, then compare this "numpy simulated" result to the PyTorch calculated one.
quantized_model_out1_int8 = np.squeeze(features['features.0.0'].int_repr().numpy())
print(quantized_model_out1_int8.shape)
print(quantized_model_out1_int8)
out1_np = QuantizedConv2D(input_scale, input_zero, input_quantized, output_scale, output_zero, weight_scale, weight_zero, weight_quantized, bias, 3, 2, 1, 112)
np.save("out1_np.npy",out1_np)
for i in range(quantized_model_out1_int8.shape[0]):
for j in range(quantized_model_out1_int8.shape[1]):
for k in range(quantized_model_out1_int8.shape[2]):
if(out1_np[i][j][k] < 0):
out1_np[i][j][k] = 0
print(out1_np)
flag = np.zeros(quantized_model_out1_int8.shape)
for i in range(quantized_model_out1_int8.shape[0]):
for j in range(quantized_model_out1_int8.shape[1]):
for k in range(quantized_model_out1_int8.shape[2]):
if(quantized_model_out1_int8[i][j][k] == out1_np[i][j][k]):
flag[i][j][k] = 1
out1_np[i][j][k] = 0
quantized_model_out1_int8[i][j][k] = 0
# Compare the simulated result to extractor fetched result, gain the total hit rate
print(flag.sum()/(112*112*32)*100,'%')
If the "numpy simulated" results are the same as the extracted one, call it a hit. Print the total hit rate, it shows that numpy gets 92% of the values right. Now the problem is, I have no idea why the rest 8% of values come out wrong.
Comparison of two outcomes:
The picture below shows the different values between Numpy one and PyTorch one, the sample channel is index[1]. The left upper corner is Numpy one, and the upright corner is PyTorch one, I have set all values that are the same between them to 0, as you can see, most of the values just have a difference of 1(This can be view as the error brought by the precision loss of fixed point arithmetics), but some have large differences, e.g. the value[1][4], 121 vs. 76 (I don't know why)
Focus on one strange value:
This code is used to replay the calculation process of the value[1][4], originally I was expecting a trial and error process could lead me to solve this problem, to get my wanted number of 76, but no matter how I tried, it didn't output 76. If you want to try this, I paste this code for your convenience.
#%% A test code to check the calculation process
weight_quantized_sample = weight_quantized[2]
M_t = input_scale * weight_scale / output_scale
ifmap_t = np.int32(input_quantized[:,1:4,7:10])
weight_t = np.int32(weight_quantized_sample)
bias_t = bias[2]
bias_q = bias_t/output_scale
res_t = 0
for ch in range(3):
ifmap_offset = ifmap_t[ch]-np.int32(input_zero)
weight_offset = weight_t[ch]-np.int32(weight_zero)
res_ch = np.multiply(ifmap_offset, weight_offset)
res_ch = res_ch.sum()
res_t = res_t + res_ch
res_mul = M_t*res_t
# for n in range(1, 30):
# res_mul = multiply(n, M_t, res_t)
res_t = round(res_mul + output_zero + bias_q)
print(res_t)
Could you help me out of this, have been stuck here for a long time.
I implemented my own version of quantized convolution and got from 99.999% to 100% hitrate (and mismatch of a single value is by 1 that I can consider to be a rounding issue). The link on the paper in the question helped a lot.
But I found that your formulas are the same as mine. So I don't know what was your issue. As I understand quantization in pytorch is hardware dependent.
Here is my code:
def my_Conv2dRelu_b2(input_q, conv_layer, output_shape):
'''
Args:
input_q: quantized tensor
conv_layer: quantized tensor
output_shape: the pre-computed shape of the result
Returns:
'''
output = np.zeros(output_shape)
# extract needed float numbers from quantized operations
weights_scale = conv_layer.weight().q_per_channel_scales()
input_scale = input_q.q_scale()
weights_zp = conv_layer.weight().q_per_channel_zero_points()
input_zp = input_q.q_zero_point()
# extract needed convolution parameters
padding = conv_layer.padding
stride = conv_layer.stride
# extract float numbers for results
output_zp = conv_layer.zero_point
output_scale = conv_layer.scale
conv_weights_int = conv_layer.weight().int_repr()
input_int = input_q.int_repr()
biases = conv_layer.bias().numpy()
for k in range(input_q.shape[0]):
for i in range(conv_weights_int.shape[0]):
output[k][i] = manual_convolution_quant(
input_int[k].numpy(),
conv_weights_int[i].numpy(),
biases[i],
padding=padding,
stride=stride,
image_zp=input_zp, image_scale=input_scale,
kernel_zp=weights_zp[i].item(), kernel_scale=weights_scale[i].item(),
result_zp=output_zp, result_scale=output_scale
)
return output
def manual_convolution_quant(image, kernel, b, padding, stride, image_zp, image_scale, kernel_zp, kernel_scale,
result_zp, result_scale):
H = image.shape[1]
W = image.shape[2]
new_H = H // stride[0]
new_W = W // stride[1]
results = np.zeros([new_H, new_W])
M = image_scale * kernel_scale / result_scale
bias = b / result_scale
paddedIm = np.pad(
image,
[(0, 0), (padding[0], padding[0]), (padding[1], padding[1])],
mode="constant",
constant_values=image_zp,
)
s = kernel.shape[1]
for i in range(new_H):
for j in range(new_W):
patch = paddedIm[
:, i * stride[0]: i * stride[0] + s, j * stride[1]: j * stride[1] + s
]
res = M * ((kernel - kernel_zp) * (patch - image_zp)).sum() + result_zp + bias
if res < 0:
res = 0
results[i, j] = round(res)
return results
Code to compare pytorch and my own version.
def calc_hit_rate(array1, array2):
good = (array1 == array2).astype(np.int).sum()
all = array1.size
return good / all
# during inference
y2 = model.conv1(y1)
y2_int = torch.int_repr(y2)
y2_int_manual = my_Conv2dRelu_b2(y1, model.conv1, y2.shape)
print(f'y2 hit rate= {calc_hit_rate(y2.int_repr().numpy(), y2_int_manual)}') #hit_rate=1.0

GP + Tensorflow training

I am trying to train a GPR model and a tensorflow model together. The training part has no issue. But for prediction using the trained model I receive a type error in a tf.placeholder op.
pred, uncp=sess.run([my, yv], feed_dict={X:xtr})
The code is similar to the 2nd example from https://gpflow.readthedocs.io/en/master/notebooks/advanced_usage.html
import numpy as np
import tensorflow as tf
import gpflow
float_type = gpflow.settings.float_type
gpflow.reset_default_graph_and_session()
def cnn_fn(x, output_dim):
out= tf.layers.dense(inputs=x, units=output_dim, activation=tf.nn.relu)
print(out)
return out
N = 150
xtr = np.random.rand(N,1)
ytr = np.sin(12*xtr) + 0.66*np.cos(25*xtr) + np.random.randn(N,1)*0.1 + 3
xtr = np.random.rand(N,28)
print(xtr.shape, ytr.shape)
nepoch=50
gp_dim=xtr.shape[1]
print(gp_dim)
minibatch_size = 16
X = tf.placeholder(tf.float32, [None, gp_dim])
Y = tf.placeholder(tf.float32, [None, 1])
with tf.variable_scope('cnn'):
f_X = tf.cast(cnn_fn(X, gp_dim), dtype=float_type)
k = gpflow.kernels.Matern52(gp_dim)
gp_model = gpflow.models.GPR(f_X, tf.cast(Y, dtype=float_type), k)
loss = -gp_model.likelihood_tensor
m, v = gp_model._build_predict(f_X)
my, yv = gp_model.likelihood.predict_mean_and_var(m, v)
with tf.variable_scope('adam'):
opt_step = tf.train.AdamOptimizer(0.001).minimize(loss)
tf_vars = tf.get_collection(tf.GraphKeys.GLOBAL_VARIABLES, scope='adam')
tf_vars += tf.get_collection(tf.GraphKeys.GLOBAL_VARIABLES, scope='cnn')
## initialize
sess = tf.Session()
sess.run(tf.variables_initializer(var_list=tf_vars))
gp_model.initialize(session=sess)
for i in range(nepoch):
shind=np.array(range(len(xtr)))
np.random.shuffle(shind)
for j in range(int(len(xtr)/minibatch_size)):
ind=shind[j*minibatch_size: (j+1)*minibatch_size]
sess.run(opt_step, feed_dict={X:xtr[ind], Y:ytr[ind]})
Executing the code above runs fine. But adding the following line gives an error:
pred, uncp=sess.run([my, yv], feed_dict={X:xtr})
with the following error:
<ipython-input-25-269715087df2> in <module>
----> 1 pred, uncp=sess.run([my, yv], feed_dict={X:xtr})
[...]
InvalidArgumentError: You must feed a value for placeholder tensor 'Placeholder_1' with dtype float and shape [?,1]
[[node Placeholder_1 (defined at <ipython-input-24-39ccf45cd248>:2) = Placeholder[dtype=DT_FLOAT, shape=[?,1], _device="/job:localhost/replica:0/task:0/device:GPU:0"]()]]
The reason your code fails is because you are not actually feeding in a value to one of the placeholders. This is easier to spot if you actually give them names:
X = tf.placeholder(tf.float32, [None, gp_dim], name='X')
Y = tf.placeholder(tf.float32, [None, 1], name='Y')
Tensorflow requires the entire compute graph to be well-defined, and the GPR model you are using depends on both X and Y. If you run the following line, it works fine:
pred, uncp = sess.run([my, yv], feed_dict={X: xtr, Y: ytr})
Update: as user1018464 pointed out in the comment, you are using the GPR model, in which the predictions directly depend on the training data (e.g. see equations (2.22) and (2.23) on page 16 of http://www.gaussianprocess.org/gpml/chapters/RW2.pdf). Hence you will need to pass in both xtr and ytr to make predictions.
Other models such as SVGP represent the function through "inducing features", commonly "pseudo input/output" pairs that summarise the data, in which case you won't need to feed in the original input values at all (I got it wrong when I first answered).
You could set up the model as follows:
gp_model = gpflow.models.SVGP(f_X, tf.cast(Y, dtype=float_type), k, gpflow.likelihoods.Gaussian(), xtr.copy(), num_data=N)
Then pred, uncp=sess.run([my, yv], feed_dict={X:xtr}) works as expected.
If you want to predict at different points Xtest, you need to set up a separate placeholder, and reuse the cnn (note the reuse=True in the variable_scope with the same name), as in example 2 of the notebook:
Xtest = tf.placeholder(tf.float32, [None, Mnist.input_dim], name='Xtest')
with tf.variable_scope('cnn', reuse=True):
f_Xtest = tf.cast(cnn_fn(Xtest, gp_dim), dtype=float_type)
Set up the model as before using f_X, but use f_Xtest in the call to _build_predict:
m, v = gp_model._build_predict(f_Xtest)
my, yv = gp_model.likelihood.predict_mean_and_var(m, v)
Now you need to pass in both X, Y, and Xtest into the session's run():
pred, uncp = sess.run([my, yv], feed_dict={X: xtr, Y: Ytr, Xtest: xtest})
where xtest is the numpy array with points at which you want to predict.
The GPflow manages TensorFlow sessions for you and you don't need to create your own TF session, when you use GPflow alone. In your case, tf.layers.dense makes
new variables, which should be initialized and I would advise to use a session which were created by GPflow. Essentially, you need to replace these lines
## initialize
sess = tf.Session()
sess.run(tf.variables_initializer(var_list=tf_vars))
gp_model.initialize(session=sess)
with:
sess = gpflow.get_default_session()
sess.run(tf.variables_initializer(var_list=tf_vars)
or wrap your code with default session context:
with tf.Session() as session:
... build and run

How would one use Pytorch's Autograd to backprop multiple layers (matracies)?

Just in order to learn I'm trying to modify this three layer neural network from scratch to still use everything from scratch EXCEPT I want to use the Pytorch Autograd function to calculate the deltas and backprop the weights. Specifically I just want to take the mean squared error of the prediction minus y stored as 'loss' and do a 'loss.backwards()' on it.
I made a successful linear regression model with Autograd but it's not working here
I've tried putting the layers and the weights inside and outside the for loop. I've done tinkering with trying to get everything to 'require_grad'.
As is, this is my current error with Python3.6
Traceback (most recent call last):
File "nn_scratch_trash.py", line 96, in <module>
layer_2.data -= layer_2.grad.data
AttributeError: 'NoneType' object has no attribute 'data'
The code...
import numpy as np
import random
import torch
np.random.seed(111)
X = np.random.randn(6,4)
y = np.random.randn(6,1)
def zeround(x):
x = 1 / ( 1 + (np.exp(-x)))
x = np.around(x)
return x
# Create Random X and y arrays of random 1s and 0s with the zeround function I made.
y = zeround(y)
X = zeround(X)
X, y = V(X), V(y)
# randomly initialize our weights
weight_0 = V(np.random.randn(4,6), requires_grad=True) - 1
weight_1 = V(np.random.randn(6,7), requires_grad=True) - 1
weight_2 = V(np.random.randn(7,1), requires_grad=True) - 1
# Feed forward through layers 0, 1, and 2
layer_1 = V(((X # weight_0)), requires_grad=True)
layer_2 = V(((layer_1 # weight_1)), requires_grad=True)
layer_3 = V(((layer_2 # weight_2)), requires_grad=True)
print((layer_1))
print((layer_2))
print((layer_3))
print(y)
for j in range(10000):
# torch backprop
loss = ((layer_3 - y))
loss.backward()
print(weight_2.requires_grad)
# UPDATE WEIGHTS
weight_0.data -= weight_0.grad.data
weight_1.data -= weight_1.grad.data
weight_2.data -= weight_2.grad.data
weight_0.grad.data.zero_()
weight_1.grad.data.zero_()
weight_2.zero_()
if j == 9999:
print (loss.item())

Kalman Filter (pykalman): Value for obs_covariance and model without intercept

I am looking at the KalmanFilter from pykalman shown in examples:
pykalman documentation
Example 1
Example 2
and I am wondering
observation_covariance=100,
vs
observation_covariance=1,
the documentation states
observation_covariance R: e(t)^2 ~ Gaussian (0, R)
How should the value be set here correctly?
Additionally, is it possible to apply the Kalman filter without intercept in the above module?
The observation covariance shows how much error you assume to be in your input data. Kalman filter works fine on normally distributed data. Under this assumption you can use the 3-Sigma rule to calculate the covariance (in this case the variance) of your observation based on the maximum error in the observation.
The values in your question can be interpreted as follows:
Example 1
observation_covariance = 100
sigma = sqrt(observation_covariance) = 10
max_error = 3*sigma = 30
Example 2
observation_covariance = 1
sigma = sqrt(observation_covariance) = 1
max_error = 3*sigma = 3
So you need to choose the value based on your observation data. The more accurate the observation, the smaller the observation covariance.
Another point: you can tune your filter by manipulating the covariance, but I think it's not a good idea. The higher the observation covariance value the weaker impact a new observation has on the filter state.
Sorry, I did not understand the second part of your question (about the Kalman Filter without intercept). Could you please explain what you mean?
You are trying to use a regression model and both intercept and slope belong to it.
---------------------------
UPDATE
I prepared some code and plots to answer your questions in details. I used EWC and EWA historical data to stay close to the original article.
First of all here is the code (pretty the same one as in the examples above but with a different notation)
from pykalman import KalmanFilter
import numpy as np
import matplotlib.pyplot as plt
# reading data (quick and dirty)
Datum=[]
EWA=[]
EWC=[]
for line in open('data/dataset.csv'):
f1, f2, f3 = line.split(';')
Datum.append(f1)
EWA.append(float(f2))
EWC.append(float(f3))
n = len(Datum)
# Filter Configuration
# both slope and intercept have to be estimated
# transition_matrix
F = np.eye(2) # identity matrix because x_(k+1) = x_(k) + noise
# observation_matrix
# H_k = [EWA_k 1]
H = np.vstack([np.matrix(EWA), np.ones((1, n))]).T[:, np.newaxis]
# transition_covariance
Q = [[1e-4, 0],
[ 0, 1e-4]]
# observation_covariance
R = 1 # max error = 3
# initial_state_mean
X0 = [0,
0]
# initial_state_covariance
P0 = [[ 1, 0],
[ 0, 1]]
# Kalman-Filter initialization
kf = KalmanFilter(n_dim_obs=1, n_dim_state=2,
transition_matrices = F,
observation_matrices = H,
transition_covariance = Q,
observation_covariance = R,
initial_state_mean = X0,
initial_state_covariance = P0)
# Filtering
state_means, state_covs = kf.filter(EWC)
# Restore EWC based on EWA and estimated parameters
EWC_restored = np.multiply(EWA, state_means[:, 0]) + state_means[:, 1]
# Plots
plt.figure(1)
ax1 = plt.subplot(211)
plt.plot(state_means[:, 0], label="Slope")
plt.grid()
plt.legend(loc="upper left")
ax2 = plt.subplot(212)
plt.plot(state_means[:, 1], label="Intercept")
plt.grid()
plt.legend(loc="upper left")
# check the result
plt.figure(2)
plt.plot(EWC, label="EWC original")
plt.plot(EWC_restored, label="EWC restored")
plt.grid()
plt.legend(loc="upper left")
plt.show()
I could not retrieve data using pandas, so I downloaded them and read from the file.
Here you can see the estimated slope and intercept:
To test the estimated data I restored the EWC value from the EWA using the estimated parameters:
About the observation covariance value
By varying the observation covariance value you tell the Filter how accurate the input data is (normally you just describe your confidence in the observation using some datasheets or your knowledge about the system).
Here are estimated parameters and the restored EWC values using different observation covariance values:
You can see the filter follows the original function better with a bigger confidence in observation (smaller R). If the confidence is low (bigger R) the filter leaves the initial estimate (slope = 0, intercept = 0) very slowly and the restored function is far away from the original one.
About the frozen intercept
If you want to freeze the intercept for some reason, you need to change the whole model and all filter parameters.
In the normal case we had:
x = [slope; intercept] #estimation state
H = [EWA 1] #observation matrix
z = [EWC] #observation
Now we have:
x = [slope] #estimation state
H = [EWA] #observation matrix
z = [EWC-const_intercept] #observation
Results:
Here is the code:
from pykalman import KalmanFilter
import numpy as np
import matplotlib.pyplot as plt
# only slope has to be estimated (it will be manipulated by the constant intercept) - mathematically incorrect!
const_intercept = 10
# reading data (quick and dirty)
Datum=[]
EWA=[]
EWC=[]
for line in open('data/dataset.csv'):
f1, f2, f3 = line.split(';')
Datum.append(f1)
EWA.append(float(f2))
EWC.append(float(f3))
n = len(Datum)
# Filter Configuration
# transition_matrix
F = 1 # identity matrix because x_(k+1) = x_(k) + noise
# observation_matrix
# H_k = [EWA_k]
H = np.matrix(EWA).T[:, np.newaxis]
# transition_covariance
Q = 1e-4
# observation_covariance
R = 1 # max error = 3
# initial_state_mean
X0 = 0
# initial_state_covariance
P0 = 1
# Kalman-Filter initialization
kf = KalmanFilter(n_dim_obs=1, n_dim_state=1,
transition_matrices = F,
observation_matrices = H,
transition_covariance = Q,
observation_covariance = R,
initial_state_mean = X0,
initial_state_covariance = P0)
# Creating the observation based on EWC and the constant intercept
z = EWC[:] # copy the list (not just assign the reference!)
z[:] = [x - const_intercept for x in z]
# Filtering
state_means, state_covs = kf.filter(z) # the estimation for the EWC data minus constant intercept
# Restore EWC based on EWA and estimated parameters
EWC_restored = np.multiply(EWA, state_means[:, 0]) + const_intercept
# Plots
plt.figure(1)
ax1 = plt.subplot(211)
plt.plot(state_means[:, 0], label="Slope")
plt.grid()
plt.legend(loc="upper left")
ax2 = plt.subplot(212)
plt.plot(const_intercept*np.ones((n, 1)), label="Intercept")
plt.grid()
plt.legend(loc="upper left")
# check the result
plt.figure(2)
plt.plot(EWC, label="EWC original")
plt.plot(EWC_restored, label="EWC restored")
plt.grid()
plt.legend(loc="upper left")
plt.show()