def main():
# Set the hyper-parameters
learning_rate = 0.001
batch_size = 100
epoches = 11
train_loader, test_loader = MNIST_dataset(batch_size)
# Build the model
lenet = Lenet()
# Set the loss function & optimizer
criterion = nn.CrossEntropyLoss()
optimizer = optim.SGD(lenet.parameters(), lr=learning_rate)
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
# Train & Test the model
training_loss = train_and_test(device,
epoches,
train_loader,
test_loader,
optimizer,
criterion,
model=lenet)
plot_loss(training_loss)
def active_learner(device, seed):
torch.backends.cudnn.deterministic = True
torch.manual_seed(seed)
torch.cuda.manual_seed(seed)
np.random.seed(seed)
random.seed(seed)
model = Lenet().to(device)
optimizer = torch.optim.Adam(model.parameters(),
lr=0.001,
amsgrad=True)
return model, optimizer, None
def main():
mnist = input_data.read_data_sets("MNIST_data/", reshape=False)
X_test, y_test = mnist.test.images, mnist.test.labels
assert (len(X_test) == len(y_test))
# Pad images with 0s
X_test = np.pad(X_test, ((0, 0), (2, 2), (2, 2), (0, 0)), 'constant')
answer = 'y'
# PLACEHOLDER FOR FEEDING INPUT DATA
X = tf.placeholder(tf.float32, shape=(None, 32, 32, 1), name="X")
net = Lenet(X)
saver = tf.train.Saver()
with tf.Session() as sess:
saver.restore(sess, cfg.FINAL_PATH)
while answer != 'n':
index = np.random.randint(0, len(X_test))
Z = net.output.eval(
feed_dict={X: X_test[index].reshape(1, 32, 32, 1)})
y_pred = np.argmax(Z, axis=1)
print("All of the classes :", Z)
print("Predicted class:", y_pred)
print("Actual class: ", y_test[index])
print()
answer = input("Do you want to predict one more number?(y/n) ")
def main():
mnist = input_data.read_data_sets("MNIST_data/", one_hot=True)
sess = tf.Session()
batch_size = cfg.BATCH_SIZE
parameter_path = cfg.PARAMETER_FILE
lenet = Lenet()
max_iter = cfg.MAX_ITER
saver = tf.train.Saver()
if os.path.exists(parameter_path):
saver.restore(parameter_path)
else:
sess.run(tf.initialize_all_variables())
for i in range(max_iter):
batch = mnist.train.next_batch(50)
if i % 100 == 0:
train_accuracy = sess.run(lenet.train_accuracy,
feed_dict={
lenet.raw_input_image: batch[0],
lenet.raw_input_label: batch[1]
})
print("step %d, training accuracy %g" % (i, train_accuracy))
sess.run(lenet.train_op,
feed_dict={
lenet.raw_input_image: batch[0],
lenet.raw_input_label: batch[1]
})
save_path = saver.save(sess, parameter_path)
def human_cnn(self):
human_cnn = Lenet()
human_cnn.load_state_dict(
torch.load(
'utils/cifar10_2class/trained_models/humancnn/lenet_weights_0.5_normalisation',
map_location=self.device))
human_cnn.to(self.device)
return human_cnn
def main():
mnist = input_data.read_data_sets("MNIST_data/", one_hot=True)
test_images = mnist.test.images
test_labels = mnist.test.labels
sess = tf.Session()
batch_size = cfg.BATCH_SIZE
parameter_path = cfg.PARAMETER_FILE
lenet = Lenet()
max_iter = cfg.MAX_ITER
saver = tf.train.Saver()
sess.run(tf.initialize_all_variables())
tf.summary.scalar("loss", lenet.loss)
summary_op = tf.summary.merge_all()
train_summary_writer = tf.summary.FileWriter("logs", sess.graph)
for i in range(max_iter):
batch = mnist.train.next_batch(batch_size)
if i % 100 == 0:
train_accuracy, summary = sess.run(
[lenet.train_accuracy, summary_op],
feed_dict={
lenet.raw_input_image: batch[0],
lenet.raw_input_label: batch[1]
})
train_summary_writer.add_summary(summary)
print("step %d, training accuracy %g" % (i, train_accuracy))
if i % 500 == 0:
test_accuracy = sess.run(lenet.train_accuracy,
feed_dict={
lenet.raw_input_image: test_images,
lenet.raw_input_label: test_labels
})
print("\n")
print("step %d, test accuracy %g" % (i, test_accuracy))
print("\n")
sess.run(lenet.train_op,
feed_dict={
lenet.raw_input_image: batch[0],
lenet.raw_input_label: batch[1]
})
saver.save(sess, parameter_path)
print("saved model")
def main():
mnist = input_data.read_data_sets("MNIST_data/", one_hot=True)
sess = tf.Session()
batch_size = cfg.BATCH_SIZE
parameter_path = cfg.PARAMETER_FILE
lenet = Lenet()
max_iter = cfg.MAX_ITER
saver = tf.train.Saver()
if os.path.exists(parameter_path):
saver.restore(parameter_path)
else:
sess.run(tf.initialize_all_variables())
temp_step = 10
result_step = np.arange(0, temp_step*100, 100)
result_acc = np.zeros(temp_step)
result_loss = np.zeros(temp_step)
result_test = np.zeros(temp_step)
for i in range(temp_step*100+1):
batch = mnist.train.next_batch(50)
if i % 100 == 0:
r = int(i / 100 - 1)
result_acc[r] = sess.run(lenet.train_accuracy,feed_dict={
lenet.raw_input_image: batch[0], lenet.raw_input_label: batch[1]
})
result_loss[r] = sess.run(lenet.loss, feed_dict={
lenet.raw_input_image: batch[0], lenet.raw_input_label: batch[1]
})
print("step %d, training accuracy %g, training loss %g" % (i, result_acc[r], result_loss[r]))
result_test[r] = sess.run(lenet.train_accuracy, feed_dict={
lenet.raw_input_image: mnist.test.images, lenet.raw_input_label: mnist.test.labels
})
print("test accuracy %g" % (result_test[r]))
sess.run(lenet.train_op,feed_dict={lenet.raw_input_image: batch[0],lenet.raw_input_label: batch[1]})
save_path = saver.save(sess, parameter_path)
plt.plot(result_step, result_acc, label='training accuracy')
plt.plot(result_step, result_test, label='test accuracy')
plt.title('LeNet')
plt.xlabel('step')
plt.ylabel('accuracy')
plt.legend()
plt.show()
def main():
image = Image.open("./img/prototype3.tiff")
image = np.array(image)
# PLACEHOLDERS FOR FEEDING INPUT DATA
X = tf.placeholder(tf.float32, shape=(None, 32, 32, 1), name="X")
net = Lenet(X)
saver = tf.train.Saver()
with tf.Session() as sess:
saver.restore(sess, cfg.FINAL_PATH)
Z = net.output.eval(feed_dict={X: image.reshape(1, 32, 32, 1)})
y_pred = np.argmax(Z, axis=1)
print("tutte le classi:", Z)
print("Predicted class:", y_pred)
def main():
# 将obs的训练数据拷贝到modelarts
mox.file.copy_parallel(src_url="obs://canncamps-hw38939615/MNIST_data/",
dst_url="MNIST_data")
mnist = input_data.read_data_sets("MNIST_data/", one_hot=True)
config = tf.ConfigProto()
custom_op = config.graph_options.rewrite_options.custom_optimizers.add()
custom_op.name = "NpuOptimizer"
config.graph_options.rewrite_options.remapping = RewriterConfig.OFF # 必须显式关闭remap
sess = tf.Session(config=config)
batch_size = cfg.BATCH_SIZE
parameter_path = cfg.PARAMETER_FILE
lenet = Lenet()
max_iter = cfg.MAX_ITER
saver = tf.train.Saver()
if os.path.exists(parameter_path):
saver.restore(parameter_path)
else:
sess.run(tf.initialize_all_variables())
for i in range(max_iter):
batch = mnist.train.next_batch(batch_size)
if i % 100 == 0:
train_accuracy, train_loss = sess.run(
[lenet.train_accuracy, lenet.loss],
feed_dict={
lenet.raw_input_image: batch[0],
lenet.raw_input_label: batch[1]
})
print("step %d, training accuracy %g, loss is %g" %
(i, train_accuracy, train_loss))
sess.run(lenet.train_op,
feed_dict={
lenet.raw_input_image: batch[0],
lenet.raw_input_label: batch[1]
})
save_path = saver.save(sess, parameter_path)
print("save model in {}".format(save_path))
# 将训练好的权重拷回到obs
mox.file.copy_parallel(src_url="checkpoint/",
dst_url="obs://canncamps-hw38939615/ckpt")
def main():
for i in range(10):
# i = 3
tf.reset_default_graph()
X = tf.Variable(tf.random_uniform(shape=(1, 32, 32, 1),
minval=0,
maxval=0.5),
name="X")
net = Lenet(X, is_trainable=False)
with tf.name_scope("am"):
am = (tf.log(net.output[:, i]) -
0.5 * tf.norm(X, ord=2, name="l2-norm"))
with tf.name_scope("optimize-am"):
optimizer = tf.train.AdamOptimizer(learning_rate=cfg.LEARNING_RATE)
training_op = optimizer.minimize(
-am) # minimizing(-cost) equals to maximize(cost)
saver = tf.train.Saver(
tf.get_collection(tf.GraphKeys.GLOBAL_VARIABLES, scope="LeNet"))
with tf.Session() as sess:
tf.global_variables_initializer().run()
saver.restore(sess, cfg.FINAL_PATH)
print("Optimizing...")
print()
for epoch in range(5000):
sess.run(training_op)
print("Epoch:", epoch, " Loss:", am.eval())
print(' Classes Probability:', net.output.eval())
img = sess.run(tf.squeeze(
X)) # -tf.reduce_min(X))/(tf.reduce_max(X)-tf.reduce_min(X))
img = Image.fromarray(img, "F")
img.save("./img/prototype{}.tiff".format(i))
print("Prototype saved as image")
def test(self):
(x_train, y_train), (x_test, y_test) = keras.datasets.mnist.load_data()
x_train = np.array(x_train) / 255
x_test = np.array(x_test) / 255
y_train = keras.utils.to_categorical(y_train)
y_test = keras.utils.to_categorical(y_test)
lenet = Lenet()
lenet.Fit(x_train, y_train, 10, 128)
score = lenet.Evaluate(x_test, y_test)
print('Test loss:', score[0])
print('Test accuracy:', score[1])
y_pred = lenet.Predict(x_test)
y_test = np.argmax(y_test, axis=1)
y_pred = np.argmax(y_pred, axis=1)
print()
print(lenet.ConfusionMatrix(y_test, y_pred))
else:
trainXX = trainXX.reshape(trainXX.shape[0], 28, 28, 1)
# data = dataset.data.astype("float") / 255.0
(trainX ,testX, trainY, testY)=train_test_split( trainXX, trainYY, test_size = 0.25)
lb = LabelBinarizer()
trainY = lb.fit_transform(trainY)
testY = lb.fit_transform(testY)
model = Lenet.load(width=28, height=28, depth=1, classes=10)
# model.add(Dense(256, input_shape=(784,), activation="sigmoid"))
# model.add(Dense(128, activation="sigmoid"))
# model.add(Dense(10, activation="softmax"))
print("[INFO] training network.....")
sgd = SGD(0.01)
model.compile(loss="categorical_crossentropy", optimizer=sgd, metrics=["accuracy"])
H = model.fit(trainX, trainY, validation_data=(testX, testY), epochs=20, batch_size=128)
print("[INFO] evaluating network.....")
predictions = model.predict(testX,batch_size=128)
print(classification_report(testY.argmax(axis=1), predictions.argmax(axis=1), target_names=[str(x) for x in lb.classes_]))
"""
将保存了训练时伪量化信息的checkpoint文件转换成freeze pb文件
"""
import tensorflow as tf
import config as cfg
from lenet import Lenet
from tensorflow.python.framework import graph_util
# os.environ['CUDA_VISIBLE_DEVICES'] = '1'
with tf.Session() as sess:
le_net = Lenet(False)
saver = tf.train.Saver(
) # 不可以导入train graph，需要重新创建一个graph，然后将train graph图中的参数来填充该图
saver.restore(sess, cfg.PARAMETER_FILE)
frozen_graph_def = graph_util.convert_variables_to_constants(
sess, sess.graph_def, ['predictions'])
tf.io.write_graph(frozen_graph_def,
"pb_model",
"freeze_eval_graph.pb",
as_text=False)
def __init__(self):
self.lenet = Lenet()
self.sess = tf.Session()
self.parameter_path = cfg.PARAMETER_FILE
self.saver = tf.train.Saver()
mnist = input_data.read_data_sets("MNIST_data/", one_hot=True)
x_train, y_train = mnist.train.images, mnist.train.labels
x_val, y_val = mnist.validation.images, mnist.validation.labels
x_test, y_test = mnist.test.images, mnist.test.labels
# Training
with tf.Graph().as_default():
session_conf = tf.ConfigProto(
allow_soft_placement=FLAGS.allow_soft_placement,
log_device_placement=FLAGS.log_device_placement
)
sess = tf.Session(config=session_conf)
with sess.as_default():
lenet = Lenet(dropout_keep_prob=FLAGS.dropout_keep_prob,
learning_rate=FLAGS.learning_rate,
l2_reg_lambda=FLAGS.l2_reg_lambda)
saver = tf.train.Saver(tf.global_variables())
if os.path.exists(FLAGS.parameter_file):
saver.restore(sess, FLAGS.paramter_file)
else:
sess.run(tf.global_variables_initializer())
for i in range(FLAGS.epoch):
batch = mnist.train.next_batch(FLAGS.batch_size)
if i % 500 == 0:
dev_accuracy = sess.run(lenet.train_accuracy, feed_dict={
lenet.raw_input_image: x_val, lenet.raw_input_label: y_val
})
def transform_img(image, is_plotted=False):
resized = tf.image.resize_images(image, (28, 28),
tf.image.ResizeMethod.AREA)
resized = resized.numpy().reshape((28, 28))
if is_plotted:
import matplotlib.pyplot as plt
plt.imshow(resized, cmap='gray')
plt.show()
resized = resized / 255
resized = resized.reshape((1, 28, 28, 1))
return resized
lenet = Lenet(20, 64, tf.train.AdamOptimizer(learning_rate=0.001),
tf.losses.softmax_cross_entropy)
lenet.load_model()
# mouse callback function
def draw_circle(event, x, y, _, __):
global start_point, end_point, drawing
if event == cv2.EVENT_LBUTTONDOWN:
drawing = True
start_point = (x, y)
elif event == cv2.EVENT_MOUSEMOVE:
if drawing:
end_point = (x, y)
cv2.line(img, start_point, end_point, 255, 20)
def main():
mnist = input_data.read_data_sets("MNIST_data/", one_hot=True)
test_images = mnist.test.images
test_labels = mnist.test.labels
sess = tf.Session()
batch_size = cfg.BATCH_SIZE
parameter_path = cfg.PARAMETER_FILE
lenet = Lenet()
max_iter = cfg.MAX_ITER
variables = get_variables_to_restore()
save_vars = [
variable for variable in variables
if not re.search("Adam", variable.name)
]
saver = tf.train.Saver(save_vars)
sess.run(tf.initialize_all_variables())
tf.summary.scalar("loss", lenet.loss)
summary_op = tf.summary.merge_all()
train_summary_writer = tf.summary.FileWriter("logs", sess.graph)
for i in range(max_iter):
batch = mnist.train.next_batch(batch_size)
if i % 100 == 0:
train_accuracy, summary = sess.run(
[lenet.train_accuracy, summary_op],
feed_dict={
lenet.raw_input_image: batch[0],
lenet.raw_input_label: batch[1]
})
train_summary_writer.add_summary(summary)
print("step %d, training accuracy %g" % (i, train_accuracy))
if i % 500 == 0:
test_accuracy = sess.run(lenet.train_accuracy,
feed_dict={
lenet.raw_input_image: test_images,
lenet.raw_input_label: test_labels
})
print("\n")
print("step %d, test accuracy %g" % (i, test_accuracy))
print("\n")
sess.run(lenet.train_op,
feed_dict={
lenet.raw_input_image: batch[0],
lenet.raw_input_label: batch[1]
})
saver.save(sess, parameter_path)
print("saved model")
# 保存为saved_Model
# Export checkpoint to SavedModel
builder = tf.saved_model.builder.SavedModelBuilder("pb_model")
inputs = {
"inputs": tf.saved_model.utils.build_tensor_info(lenet.raw_input_image)
}
outputs = {
"predictions":
tf.saved_model.utils.build_tensor_info(lenet.predictions)
}
prediction_signature = tf.saved_model.signature_def_utils.build_signature_def(
inputs=inputs,
outputs=outputs,
method_name=tf.saved_model.signature_constants.PREDICT_METHOD_NAME)
legacy_init_op = tf.group(tf.tables_initializer(), name="legacy_init_op")
builder.add_meta_graph_and_variables(
sess, [tf.saved_model.tag_constants.SERVING],
signature_def_map={"serving_default": prediction_signature},
legacy_init_op=legacy_init_op,
saver=saver)
builder.save()
X_test = X_test[:, :, :, None]
train_data = (X_train, y_train)
test_data = (X_test, y_test)
train_data = (train_data[0],
keras.utils.to_categorical(train_data[1],
num_classes=CLASS_NUM))
test_data = (test_data[0],
keras.utils.to_categorical(test_data[1],
num_classes=CLASS_NUM))
# get GAN discriminator score for training set
d_pret = d.predict(X_train, verbose=1)
d_pret = d_pret.reshape((60000, ))
#d_pret = np.sort(d_pret)
disc_scores = preprocessing.scale(d_pret) + 0.5
# Create a classifier
classifier_daal = Lenet(session, scope='daal_lenet')
#do the experiment many times with different random_seeds
daal_all = []
for iteration in range(ACTIVE_LEARNING_ITERS):
print "Active Learning Iteration:", iteration
daal_accuracies = []
#shuffle the training data
images, targets = train_data[0].copy(), train_data[1].copy(
) #TODO better not to copy and play with indices, instead. whatever!
scores = disc_scores[:]
assert len(scores) == train_data[0].shape[0]
rng_state = np.random.get_state()
np.random.shuffle(images)
np.random.set_state(rng_state)
np.random.shuffle(targets)
np.random.set_state(rng_state)
def main():
now = datetime.utcnow().strftime("%Y%m%d%H%M%S")
root_logdir = "tf_logs"
logdir = "{}/run-{}/".format(root_logdir, now)
mnist = input_data.read_data_sets("MNIST_data/", reshape=False)
X_train, y_train = mnist.train.images, mnist.train.labels
X_validation, y_validation = mnist.validation.images, mnist.validation.labels
X_test, y_test = mnist.test.images, mnist.test.labels
assert (len(X_train) == len(y_train))
assert (len(X_validation) == len(y_validation))
assert (len(X_test) == len(y_test))
print()
print("Image Shape: {}".format(X_train[0].shape))
print()
print("Training Set: {} samples".format(len(X_train)))
print("Validation Set: {} samples".format(len(X_validation)))
print("Test Set: {} samples".format(len(X_test)))
#The MNIST data that TensorFlow pre-loads comes as 28x28x1 images.
#However, the LeNet architecture only accepts 32x32xC images, where C is the number of color channels.
# Pad images with 0s
X_train = np.pad(X_train, ((0, 0), (2, 2), (2, 2), (0, 0)), 'constant')
X_validation = np.pad(X_validation, ((0, 0), (2, 2), (2, 2), (0, 0)),
'constant')
X_test = np.pad(X_test, ((0, 0), (2, 2), (2, 2), (0, 0)), 'constant')
print("Updated Image Shape: {}".format(X_train[0].shape))
#Shuffle the training data
X_train, y_train = shuffle(X_train, y_train)
X = tf.placeholder(tf.float32, shape=(None, 32, 32, 1), name="X")
y = tf.placeholder(tf.int32, shape=(None), name="y")
one_hot_y = tf.one_hot(y, 10)
net = Lenet(X, is_trainable=True)
with tf.name_scope("loss"):
cross_entropy = tf.nn.softmax_cross_entropy_with_logits(
labels=one_hot_y, logits=net.output)
loss_operation = tf.reduce_mean(cross_entropy, name="loss")
with tf.name_scope("train"):
optimizer = tf.train.AdamOptimizer(learning_rate=cfg.LEARNING_RATE)
training_op = optimizer.minimize(loss_operation)
with tf.name_scope("eval"):
correct = tf.equal(tf.argmax(net.output, 1), tf.argmax(one_hot_y, 1))
accuracy = tf.reduce_mean(tf.cast(correct, tf.float32))
init = tf.global_variables_initializer()
saver = tf.train.Saver()
cross_entropy_summary = tf.summary.scalar('cross_entropy', loss_operation)
acc_train_summary = tf.summary.scalar('training_accuracy', accuracy)
acc_val_summary = tf.summary.scalar('validation_accuracy', accuracy)
file_writer = tf.summary.FileWriter(logdir, tf.get_default_graph())
num_examples = len(X_train)
num_batches = num_examples // cfg.BATCH_SIZE
with tf.Session() as sess:
init.run()
print("Training...")
print()
for epoch in range(cfg.EPOCHS):
X_train, y_train = shuffle(X_train, y_train)
batch_index = 0
for offset in range(0, num_examples, cfg.BATCH_SIZE):
end = offset + cfg.BATCH_SIZE
X_batch, y_batch = X_train[offset:end], y_train[offset:end]
if batch_index % 10 == 0:
cross_entropy_str = cross_entropy_summary.eval(feed_dict={
X: X_batch,
y: y_batch
})
step = epoch * num_batches + batch_index
file_writer.add_summary(cross_entropy_str, step)
sess.run(training_op, feed_dict={X: X_batch, y: y_batch})
batch_index += 1
acc_train_str = acc_train_summary.eval(feed_dict={
X: X_train,
y: y_train
})
acc_val_str = acc_val_summary.eval(feed_dict={
X: X_validation,
y: y_validation
})
file_writer.add_summary(acc_train_str, epoch)
file_writer.add_summary(acc_val_str, epoch)
print("Epoch:", epoch)
save_path = saver.save(sess, cfg.INTERMEDIATE_PATH)
save_path = saver.save(sess, cfg.FINAL_PATH)
print("Model Saved")
file_writer.close()
with tf.Session() as sess:
saver.restore(sess, save_path)
acc_test = accuracy.eval(feed_dict={X: X_test, y: y_test})
print("Test Accuracy = {:.3f}".format(acc_test))
#saver = tf.train.import_meta_graph(MODEL_PATH+'/model-100.meta')
print MODEL_PATH
with tf.Session(graph=new_graph) as session:
# Create and Restore pre-trained GAN
real_data = tf.placeholder(tf.float32, shape=[128, OUTPUT_DIM])
disc_real = Discriminator(real_data)
var_list = [v for v in tf.global_variables() if "Discriminator" in v.name ]
saver = tf.train.Saver(var_list)
saver.restore(session, tf.train.latest_checkpoint(MODEL_PATH))
# get GAN discriminator score for training set
train_data, dev_data, test_data = get_mnist()
disc_scores = run_all(session, disc_real, real_data, train_data[0])
disc_scores -= min(disc_scores)
disc_scores /= max(disc_scores)
# Create a classifier
classifier_entropy = Lenet(None,scope='entropy_lenet')
classifier_daal = Lenet(None,scope='daal_lenet')
var_list = [v for v in tf.global_variables() if "target" in v.name ]
print "*************"
print var_list
var_list = [v for v in tf.global_variables() if "lenet" in v.name ]
print "*************"
for v in var_list:
print v
train_data = (train_data[0], keras.utils.to_categorical(train_data[1], num_classes=CLASS_NUM))
test_data = (test_data[0], keras.utils.to_categorical(test_data[1], num_classes=CLASS_NUM))
#do the experiment many times with different random_seeds
daal_all= []
entropy_all = []
for iteration in range(ACTIVE_LEARNING_ITERS):
print "Active Learning Iteration:", iteration
image_datasets = {
x: datasets.ImageFolder(os.path.join(data_dir, x), data_transforms[x])
for x in ['train', 'val']
}
dataloaders = {
x: torch.utils.data.DataLoader(image_datasets[x],
batch_size=20,
shuffle=True,
num_workers=4)
for x in ['train', 'val']
}
dataset_sizes = {x: len(image_datasets[x]) for x in ['train', 'val']}
class_names = image_datasets['train'].classes
print("class_names", class_names)
model_ft = Lenet()
model_ft.conv1 = nn.Conv2d(3, 6, 3) # 这里是彩色图像
num_ftrs = model_ft.fc3.in_features
# Alternatively, it can be generalized to nn.Linear(num_ftrs, len(class_names)).
model_ft.fc = nn.Linear(num_ftrs, len(class_names))
if torch.cuda.device_count() > 1:
print("Let's use", torch.cuda.device_count(), "GPUs!")
# dim = 0 [30, xxx] -> [10, ...], [10, ...], [10, ...] on 3 GPUs
model_ft = nn.DataParallel(model_ft)
model_ft = model_ft.to(device)
criterion = nn.CrossEntropyLoss()
# Observe that all parameters are being optimized
import tensorflow as tf
from tensorflow.contrib.learn.python.learn.datasets.mnist import read_data_sets
from lenet import Lenet
mnist = read_data_sets("mnist_data/", one_hot=True)
batch_size = 100
lenet_part = Lenet(mu = 0, sigma = 0.3, learning_rate = 0.001)
merged = lenet_part.merged_summary
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
train_writer = tf.summary.FileWriter('./log/train', sess.graph)
test_writer = tf.summary.FileWriter('./log/test')
for i in range(3000):
batch = mnist.train.next_batch(batch_size)
_, train_acc, train_sum= sess.run([lenet_part.training_step, lenet_part.accuracy, merged],
feed_dict = {lenet_part.raw_input_image: batch[0],lenet_part.raw_input_label: batch[1]})
test_acc, test_sum = sess.run([lenet_part.accuracy, merged],
feed_dict = {lenet_part.raw_input_image: mnist.test.images,
lenet_part.raw_input_label: mnist.test.labels})
if i%10 ==0:
test_writer.add_summary(test_sum, i)
train_writer.add_summary(train_sum, i)
print('[train_acc, test_acc]: ', train_acc, test_acc)
train_writer.close()
test_writer.close()
