您好,登錄后才能下訂單哦!
將訓(xùn)練好的模型參數(shù)保存起來(lái),以便以后進(jìn)行驗(yàn)證或測(cè)試,這是我們經(jīng)常要做的事情。tf里面提供模型保存的是tf.train.Saver()模塊。
模型保存,先要?jiǎng)?chuàng)建一個(gè)Saver對(duì)象:如
saver=tf.train.Saver()
在創(chuàng)建這個(gè)Saver對(duì)象的時(shí)候,有一個(gè)參數(shù)我們經(jīng)常會(huì)用到,就是 max_to_keep 參數(shù),這個(gè)是用來(lái)設(shè)置保存模型的個(gè)數(shù),默認(rèn)為5,即 max_to_keep=5,保存最近的5個(gè)模型。如果你想每訓(xùn)練一代(epoch)就想保存一次模型,則可以將 max_to_keep設(shè)置為None或者0,如:
saver=tf.train.Saver(max_to_keep=0)
但是這樣做除了多占用硬盤,并沒有實(shí)際多大的用處,因此不推薦。
當(dāng)然,如果你只想保存最后一代的模型,則只需要將max_to_keep設(shè)置為1即可,即
saver=tf.train.Saver(max_to_keep=1)
創(chuàng)建完saver對(duì)象后,就可以保存訓(xùn)練好的模型了,如:
saver.save(sess,'ckpt/mnist.ckpt',global_step=step)
第一個(gè)參數(shù)sess,這個(gè)就不用說(shuō)了。第二個(gè)參數(shù)設(shè)定保存的路徑和名字,第三個(gè)參數(shù)將訓(xùn)練的次數(shù)作為后綴加入到模型名字中。
saver.save(sess, 'my-model', global_step=0) ==> filename: 'my-model-0'
...
saver.save(sess, 'my-model', global_step=1000) ==> filename: 'my-model-1000'
看一個(gè)mnist實(shí)例:
# -*- coding: utf-8 -*- """ Created on Sun Jun 4 10:29:48 2017 @author: Administrator """ import tensorflow as tf from tensorflow.examples.tutorials.mnist import input_data mnist = input_data.read_data_sets("MNIST_data/", one_hot=False) x = tf.placeholder(tf.float32, [None, 784]) y_=tf.placeholder(tf.int32,[None,]) dense1 = tf.layers.dense(inputs=x, units=1024, activation=tf.nn.relu, kernel_initializer=tf.truncated_normal_initializer(stddev=0.01), kernel_regularizer=tf.nn.l2_loss) dense2= tf.layers.dense(inputs=dense1, units=512, activation=tf.nn.relu, kernel_initializer=tf.truncated_normal_initializer(stddev=0.01), kernel_regularizer=tf.nn.l2_loss) logits= tf.layers.dense(inputs=dense2, units=10, activation=None, kernel_initializer=tf.truncated_normal_initializer(stddev=0.01), kernel_regularizer=tf.nn.l2_loss) loss=tf.losses.sparse_softmax_cross_entropy(labels=y_,logits=logits) train_op=tf.train.AdamOptimizer(learning_rate=0.001).minimize(loss) correct_prediction = tf.equal(tf.cast(tf.argmax(logits,1),tf.int32), y_) acc= tf.reduce_mean(tf.cast(correct_prediction, tf.float32)) sess=tf.InteractiveSession() sess.run(tf.global_variables_initializer()) saver=tf.train.Saver(max_to_keep=1) for i in range(100): batch_xs, batch_ys = mnist.train.next_batch(100) sess.run(train_op, feed_dict={x: batch_xs, y_: batch_ys}) val_loss,val_acc=sess.run([loss,acc], feed_dict={x: mnist.test.images, y_: mnist.test.labels}) print('epoch:%d, val_loss:%f, val_acc:%f'%(i,val_loss,val_acc)) saver.save(sess,'ckpt/mnist.ckpt',global_step=i+1) sess.close()
代碼中紅色部分就是保存模型的代碼,雖然我在每訓(xùn)練完一代的時(shí)候,都進(jìn)行了保存,但后一次保存的模型會(huì)覆蓋前一次的,最終只會(huì)保存最后一次。因此我們可以節(jié)省時(shí)間,將保存代碼放到循環(huán)之外(僅適用max_to_keep=1,否則還是需要放在循環(huán)內(nèi)).
在實(shí)驗(yàn)中,最后一代可能并不是驗(yàn)證精度最高的一代,因此我們并不想默認(rèn)保存最后一代,而是想保存驗(yàn)證精度最高的一代,則加個(gè)中間變量和判斷語(yǔ)句就可以了。
saver=tf.train.Saver(max_to_keep=1) max_acc=0 for i in range(100): batch_xs, batch_ys = mnist.train.next_batch(100) sess.run(train_op, feed_dict={x: batch_xs, y_: batch_ys}) val_loss,val_acc=sess.run([loss,acc], feed_dict={x: mnist.test.images, y_: mnist.test.labels}) print('epoch:%d, val_loss:%f, val_acc:%f'%(i,val_loss,val_acc)) if val_acc>max_acc: max_acc=val_acc saver.save(sess,'ckpt/mnist.ckpt',global_step=i+1) sess.close()
如果我們想保存驗(yàn)證精度最高的三代,且把每次的驗(yàn)證精度也隨之保存下來(lái),則我們可以生成一個(gè)txt文件用于保存。
saver=tf.train.Saver(max_to_keep=3) max_acc=0 f=open('ckpt/acc.txt','w') for i in range(100): batch_xs, batch_ys = mnist.train.next_batch(100) sess.run(train_op, feed_dict={x: batch_xs, y_: batch_ys}) val_loss,val_acc=sess.run([loss,acc], feed_dict={x: mnist.test.images, y_: mnist.test.labels}) print('epoch:%d, val_loss:%f, val_acc:%f'%(i,val_loss,val_acc)) f.write(str(i+1)+', val_acc: '+str(val_acc)+'\n') if val_acc>max_acc: max_acc=val_acc saver.save(sess,'ckpt/mnist.ckpt',global_step=i+1) f.close() sess.close()
模型的恢復(fù)用的是restore()函數(shù),它需要兩個(gè)參數(shù)restore(sess, save_path),save_path指的是保存的模型路徑。我們可以使用tf.train.latest_checkpoint()來(lái)自動(dòng)獲取最后一次保存的模型。如:
model_file=tf.train.latest_checkpoint('ckpt/') saver.restore(sess,model_file)
則程序后半段代碼我們可以改為:
sess=tf.InteractiveSession() sess.run(tf.global_variables_initializer()) is_train=False saver=tf.train.Saver(max_to_keep=3) #訓(xùn)練階段 if is_train: max_acc=0 f=open('ckpt/acc.txt','w') for i in range(100): batch_xs, batch_ys = mnist.train.next_batch(100) sess.run(train_op, feed_dict={x: batch_xs, y_: batch_ys}) val_loss,val_acc=sess.run([loss,acc], feed_dict={x: mnist.test.images, y_: mnist.test.labels}) print('epoch:%d, val_loss:%f, val_acc:%f'%(i,val_loss,val_acc)) f.write(str(i+1)+', val_acc: '+str(val_acc)+'\n') if val_acc>max_acc: max_acc=val_acc saver.save(sess,'ckpt/mnist.ckpt',global_step=i+1) f.close() #驗(yàn)證階段 else: model_file=tf.train.latest_checkpoint('ckpt/') saver.restore(sess,model_file) val_loss,val_acc=sess.run([loss,acc], feed_dict={x: mnist.test.images, y_: mnist.test.labels}) print('val_loss:%f, val_acc:%f'%(val_loss,val_acc)) sess.close()
標(biāo)紅的地方,就是與保存、恢復(fù)模型相關(guān)的代碼。用一個(gè)bool型變量is_train來(lái)控制訓(xùn)練和驗(yàn)證兩個(gè)階段。
整個(gè)源程序:
# -*- coding: utf-8 -*- """ Created on Sun Jun 4 10:29:48 2017 @author: Administrator """ import tensorflow as tf from tensorflow.examples.tutorials.mnist import input_data mnist = input_data.read_data_sets("MNIST_data/", one_hot=False) x = tf.placeholder(tf.float32, [None, 784]) y_=tf.placeholder(tf.int32,[None,]) dense1 = tf.layers.dense(inputs=x, units=1024, activation=tf.nn.relu, kernel_initializer=tf.truncated_normal_initializer(stddev=0.01), kernel_regularizer=tf.nn.l2_loss) dense2= tf.layers.dense(inputs=dense1, units=512, activation=tf.nn.relu, kernel_initializer=tf.truncated_normal_initializer(stddev=0.01), kernel_regularizer=tf.nn.l2_loss) logits= tf.layers.dense(inputs=dense2, units=10, activation=None, kernel_initializer=tf.truncated_normal_initializer(stddev=0.01), kernel_regularizer=tf.nn.l2_loss) loss=tf.losses.sparse_softmax_cross_entropy(labels=y_,logits=logits) train_op=tf.train.AdamOptimizer(learning_rate=0.001).minimize(loss) correct_prediction = tf.equal(tf.cast(tf.argmax(logits,1),tf.int32), y_) acc= tf.reduce_mean(tf.cast(correct_prediction, tf.float32)) sess=tf.InteractiveSession() sess.run(tf.global_variables_initializer()) is_train=True saver=tf.train.Saver(max_to_keep=3) #訓(xùn)練階段 if is_train: max_acc=0 f=open('ckpt/acc.txt','w') for i in range(100): batch_xs, batch_ys = mnist.train.next_batch(100) sess.run(train_op, feed_dict={x: batch_xs, y_: batch_ys}) val_loss,val_acc=sess.run([loss,acc], feed_dict={x: mnist.test.images, y_: mnist.test.labels}) print('epoch:%d, val_loss:%f, val_acc:%f'%(i,val_loss,val_acc)) f.write(str(i+1)+', val_acc: '+str(val_acc)+'\n') if val_acc>max_acc: max_acc=val_acc saver.save(sess,'ckpt/mnist.ckpt',global_step=i+1) f.close() #驗(yàn)證階段 else: model_file=tf.train.latest_checkpoint('ckpt/') saver.restore(sess,model_file) val_loss,val_acc=sess.run([loss,acc], feed_dict={x: mnist.test.images, y_: mnist.test.labels}) print('val_loss:%f, val_acc:%f'%(val_loss,val_acc)) sess.close()
參考文章:https://www.jb51.net/article/138779.htm
以上就是本文的全部?jī)?nèi)容,希望對(duì)大家的學(xué)習(xí)有所幫助,也希望大家多多支持億速云。
免責(zé)聲明:本站發(fā)布的內(nèi)容(圖片、視頻和文字)以原創(chuàng)、轉(zhuǎn)載和分享為主,文章觀點(diǎn)不代表本網(wǎng)站立場(chǎng),如果涉及侵權(quán)請(qǐng)聯(lián)系站長(zhǎng)郵箱:is@yisu.com進(jìn)行舉報(bào),并提供相關(guān)證據(jù),一經(jīng)查實(shí),將立刻刪除涉嫌侵權(quán)內(nèi)容。