I have a frozen inference graph stored in a .pb file, which was obtained from a trained Tensorflow model by the freeze_graph function.
Suppose, for simplicity, that I would like to change some of the sigmoid activations in the model to tanh activations (and let's not discuss whether this is a good idea).
How can this be done with access only to the frozen graph in the .pb file, and without the possibility to retrain the model?
I am aware of the Graph Editor library in tf.contrib, which should be able to do this kind of job, but I wasn't able to figure out a simple way to do this in the documentation.
The solution is to use import_graph_def:
import tensorflow as tf
sess = tf.Session()
def load_graph(frozen_graph_filename):
with tf.gfile.GFile(frozen_graph_filename, "rb") as f:
graph_def = tf.GraphDef()
graph_def.ParseFromString(f.read())
with tf.Graph().as_default() as graph:
tf.import_graph_def(graph_def, name='')
return graph
graph_model = load_graph("frozen_inference_graph.pb")
graph_model_def = graph_model.as_graph_def()
graph_new = tf.Graph()
graph_new.as_default()
my_new_tensor = # whatever
tf.import_graph_def(graph_model_def, name='', input_map={"tensor_to_replace": my_new_tensor})
#do somthing with your new graph
Here I wrote a post about it
Can you try this:
graph = load_graph(filename)
graph_def = graph.as_graph_def()
# if ReLu op is at node 161
graph_def.node[161].op="tanh"
tf.train.write_graph(graph_def, path2savfrozn, "altered_frozen.pb", False)
Please let know the if it works.
The *.pb file contains a SavedModel protocol buffer. You should be able to load it using a SavedModel loader. You can also inpsect it with the SavedModel CLI. The full documentation on SavedModels is here.
Something along these lines should work:
graph_def = tf.GraphDef()
with open('frozen_inference.pb', 'rb') as f:
graph_def.ParseFromString(f.read())
with tf.Graph().as_default() as graph:
importer.import_graph_def(graph_def, name='')
new_model = tf.GraphDef()
with tf.Session(graph=graph) as sess:
for n in sess.graph_def.node:
if n.op == 'Sigmoid':
nn = new_model.node.add()
nn.op = 'Tanh'
nn.name = n.name
for i in n.input:
nn.input.extend([i])
else:
nn = new_model.node.add()
nn.CopyFrom(n)
Related
I wonder what is the right way to use torchtext for inference.
Let's assume I've trained the model and dump all Fields with built vocabularies. It seems the next step is to use torchtext.data.Example to load one single example. Somehow I should numeralize it by using loaded Fields and create an Iterator.
I would appreciate any simple examples of using torchtext for inference.
For a trained model and vocabulary (which is part of the text field , you don't have to save the whole class) :
def read_vocab(path):
#read vocabulary pkl
import pickle
pkl_file = open(path, 'rb')
vocab = pickle.load(pkl_file)
pkl_file.close()
return vocab
def load_model_and_vocab():
import torch
import os.path
my_path = os.path.abspath(os.path.dirname(__file__))
vocab_path = os.path.join(my_path, vocab_file)
weights_path = os.path.join(my_path, WEIGHTS)
vocab = read_vocab(vocab_path)
model = classifier(vocab_size=len(vocab))
model.load_state_dict(torch.load(weights_path))
model.eval()
return model, vocab
def predict(model, vocab, sentence):
tokenized = [w.text.lower() for w in nlp(sentence)] # tokenize the sentence
indexed = [vocab.stoi[t] for t in tokenized] # convert to integer sequence
length = [len(indexed)] # compute no. of words
tensor = torch.LongTensor(indexed).to('cpu') # convert to tensor
tensor = tensor.unsqueeze(1).T # reshape in form of batch,no. of words
length_tensor = torch.LongTensor(length) # convert to tensor
prediction = model(tensor, length_tensor) # prediction
return round(1-prediction.item())
"classifier" is the class I defined for my model.
For saving the vocabulary pkl :
def save_vocab(vocab):
import pickle
output = open('vocab.pkl', 'wb')
pickle.dump(vocab, output)
output.close()
And for saving the model after training you can use :
torch.save(model.state_dict(), 'saved_weights.pt')
Tell me if it worked for you!
I want to fit a Siamese CNN with multiple inputs that are stored in my memory and no label (just an arbitrary dummy label). Therefore, I had to write my own data_generator function for using a CNN model in Keras.
My data generator is of the following form
class DataGenerator(keras.utils.Sequence):
def __init__(self, train_data, train_triplets, batch_size=32, dim=(128,128), n_channels=3, shuffle=True):
self.dim = dim
self.batch_size = batch_size
#Added
self.train_data = train_data
self.train_triplets = train_triplets
self.n_channels = n_channels
self.shuffle = shuffle
self.on_epoch_end()
def __len__(self):
'Denotes the number of batches per epoch'
n_row = self.train_triplets.shape[0]
return int(np.floor(n_row / self.batch_size))
def __getitem__(self, index):
'Generate one batch of data'
# Generate indexes of the batch
#print(index)
indexes = self.indexes[index*self.batch_size:(index+1)*self.batch_size]
# Find list of IDs
list_IDs_temp = self.train_triplets.iloc[indexes,]
# Generate data
[anchor, positive, negative] = self.__data_generation(list_IDs_temp)
y_train = np.random.randint(2, size=(1,2,self.batch_size)).T
return [anchor,positive, negative], y_train
def on_epoch_end(self):
'Updates indexes after each epoch'
n_row = self.train_triplets.shape[0]
self.indexes = np.arange(n_row)
if self.shuffle == True:
np.random.shuffle(self.indexes)
def __data_generation(self, list_IDs_temp):
'Generates data containing batch_size samples'
# anchor positive and negatives: (n_samples, *dim, n_channels)
# Initialization
anchor = np.zeros((self.batch_size,*self.dim,self.n_channels))
positive = np.zeros((self.batch_size,*self.dim,self.n_channels))
negative = np.zeros((self.batch_size,*self.dim,self.n_channels))
nrow_temp = list_IDs_temp.shape[0]
# Generate data
for i in range(nrow_temp):
list_ind = list_IDs_temp.iloc[i,]
anchor[i] = self.train_data[list_ind[0]]
positive[i] = self.train_data[list_ind[1]]
negative[i] = self.train_data[list_ind[2]]
return [anchor, positive, negative]
where train_data is a list of all images and train triplets a data frame containing image indices to create my inputs containing of a triplet of images.
Now, I want to do some data augmenting for each mini batch supplied to my CNN. I have tried to integrate the ImageDataGenarator of Keras but I couldn't implement it in my code. Is it somehow possible to do it ? I am not very experienced with python and would appreciate any help.
Does this article answer your question?
To put it in a nutshell, Kera's ImageDataGenerator lacks flexibility when it comes to personalized batch generators, and the easiest way to still use data augmentation is simply to switch to another data augmentation tool (like the albumentations library described in the previous article, but you could also use imgaug as well).
I just want to warn you that I encountered several issues with albumentations (that I described in this question on GitHub, but for now I still have had no answers), so maybe using imgaug is a better idea.
Hope this helps, good luck with your model !
I used the plot_importance to show me the importance variables. But some variables are categorical, so I did some transformation. After I transformed the type of the variables, when i plot importance features, the plot does not show me feature names. I attached my code, and the plot.
dataset = data.values
X = dataset[1:100,0:-2]
predictors=dataset[1:100,-1]
X = X.astype(str)
encoded_x = None
for i in range(0, X.shape[1]):
label_encoder = LabelEncoder()
feature = label_encoder.fit_transform(X[:,i])
feature = feature.reshape(X.shape[0], 1)
onehot_encoder = OneHotEncoder(sparse=False)
feature = onehot_encoder.fit_transform(feature)
if encoded_x is None:
encoded_x = feature
else:
encoded_x = np.concatenate((encoded_x, feature), axis=1)
print("X shape: : ", encoded_x.shape)
response='Default'
#predictors=list(data.columns.values[:-1])
# Randomly split indexes
X_train, X_test, y_train, y_test = train_test_split(encoded_x,predictors,train_size=0.7, random_state=5)
model = XGBClassifier()
model.fit(X_train, y_train)
plot_importance(model)
plt.show()
[enter image description here][1]
[1]: https://i.stack.imgur.com/M9qgY.png
This is the expected behaviour- sklearn.OneHotEncoder.transform() returns a numpy 2d array instead of the input pd.DataFrame (i assume that's the type of your dataset). So it is not a bug, but a feature. It doesn't look like there is a way to pass feature names manually in the sklearn API (it is possible to set those in xgb.Dmatrix creation in the native training API).
However, your problem is easily solvable with pd.get_dummies() instead of the LabelEncoder + OneHotEncoder combination that you have implemented. I do not know why did you choose to use it instead (it can be useful, if you need to handle also a test set but then you need to play extra tricks), but i would advise in favour of pd.get_dummies()
I wanted to know what is the correct way to save a tensorflow model that I have trained in python so that I can import it in OpenCV using the dnn module of opencv. This is my Tensorflow graph
X = tf.placeholder(tf.float32, [None,training_set.shape[1]],name = 'X')
Y = tf.placeholder(tf.float32,[None,training_labels.shape[1]], name = 'Y')
A1 = tf.contrib.layers.fully_connected(X, num_outputs = 50, activation_fn = tf.nn.relu)
A1 = tf.nn.dropout(A1, 0.8)
A2 = tf.contrib.layers.fully_connected(A1, num_outputs = 2, activation_fn = None)
cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits = A2, labels = Y))
global_step = tf.Variable(0, trainable=False)
start_learning_rate = 0.001
learning_rate = tf.train.exponential_decay(start_learning_rate, global_step, 100, 0.1, True )
optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate).minimize(cost)
As you can see it doesn't contain any variables. So my question is how should this graph be saved in Tensorflow so that it can be loaded using cv::dnn::readNetFromTensorflow. Should I save the model as .pb or .pbtxt file. And will the .pb or .pbtxt contain the graph as well as the weights or just the graph ??. How to load both the graph and the weights in OpenCV ??.
The code that belongs to OP posted link is posted here. URL may change, code renamed or vanished. Therefore I've posted the code to where it is referred by OP.
I guess a first question is how to save the graph at least to load it in TensorFlow again? Because you need to find a way to restore it. There is some way to do it:
-- Save
# Save a graph definition (once)
tf.train.write_graph(sess.graph.as_graph_def(), "", "graph.pb")
# Weights initialization
sess.run(tf.global_variables_initializer())
# Training
...
# Save a checkpoint (weights only, no graph definition)
saver = tf.train.Saver()
saver.save(sess, 'tmp.ckpt')
-- Freeze (merge graph definition with weights, remove training-only nodes)
python ~/tensorflow/tensorflow/python/tools/freeze_graph.py \
--input_graph=graph.pb \
--input_checkpoint=tmp.ckpt \
--output_graph=frozen_graph.pb \
--output_node_names="NameOfOutputNode"
Only after these steps you might load frozen_graph.pb contains both graph definition and weights using OpenCV.
I am using the TensorFlow for Poets code lab to guide me as I retrain the Inceptionv3 CNN to classify a list of images. I have successfully trained the model, and it works when i employ the given code to classify individual images. But when i try and use it on a large batch of images, then i get the GraphDef cannot be larger than 2GB. Please advise.
import pandas as pd
import os, sys
import tensorflow as tf
test_images = pd.read_csv('test_images.csv')
testid = test_images['Id']
listx= list(range(4320))
predlist=[]
output = pd.DataFrame({'Id': listx})
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
for x in listx:
path = 'test/'+str(x+1)+'.jpg'
# change this as you see fit
image_path = path
# Read in the image_data
image_data = tf.gfile.FastGFile(image_path, 'rb').read()
# Loads label file, strips off carriage return
label_lines = [line.rstrip() for line
in tf.gfile.GFile("retrained_labels.txt")]
# Unpersists graph from file
with tf.gfile.FastGFile("retrained_graph.pb", 'rb') as f:
graph_def = tf.GraphDef()
graph_def.ParseFromString(f.read())
tf.import_graph_def(graph_def, name='')
with tf.Session() as sess:
# Feed the image_data as input to the graph and get first prediction
with tf.Graph().as_default():
softmax_tensor = sess.graph.get_tensor_by_name('final_result:0')
predictions = sess.run(softmax_tensor, \
{'DecodeJpeg/contents:0': image_data})
# Sort to show labels of first prediction in order of confidence
top_k = predictions[0].argsort()[-len(predictions[0]):][::-1]
# print('the top result is' + label_lines[node_id])
flag = 0
for node_id in top_k:
while flag == 0:
human_string = label_lines[node_id]
score = predictions[0][node_id]
predlist.append(int(human_string[:3]))
print('%s' % (human_string))
flag = 1 # we only want the top prediction
output['Prediction']=predlist
output.to_csv('outputtest.csv')
One way by which this error can e solved is by placing
with tf.Graph().as_default():
after for loop.
This is the piece of code that worked for me while trying to read bulk image:
for filename in os.listdir(image_path):
with tf.Graph().as_default():
# Read in the image_data
image_data = tf.gfile.FastGFile(image_path + filename, 'rb').read()