Cannot feed value of shape (10000, 784) for Tensor about deformable_convnet

[jiachen0212]

you need to modified the main.py, such as:
def train(self):
create_dir(SAVE_DIR)
saver = tf.train.Saver(tf.global_variables(), max_to_keep = 10)
ntrain = len(self.train_data.image) # traindata : 55000 testdata : 10000
ntest = len(self.test_data.image)
nbatch = ntrain//self.batch_size
nbatch_t = ntest // self.batch_size

    for epoch in range(self.epoch):
        # shuffle start
        index = np.arange(ntrain)
        np.random.shuffle(index)
        shuffle_image = self.train_data.image[index]   # shuffle_image: 55000,784
        shuffle_label = self.train_data.label[index]

        ########################## shuffle the testdata #####################
        index_t = np.arange(ntest)
        np.random.shuffle(index_t)
        shuffle_test = self.test_data.image[index_t]   # shuffle_test: 10000,784
        shuffle_label_t = self.test_data.label[index_t]
        ######################################################################
        # shuffle end

        Train_accuracy = 0
        Test_accuracy = 0
        for batch in tqdm(range(nbatch), ascii = True, desc = "batch"):
            start = self.batch_size*batch
            end = self.batch_size*(batch+1)
            train_feed_dict = {self.image : shuffle_image[start:end], self.label : shuffle_label[start:end]}
            _, batch_accuracy = self.sess.run([self.run_train, self.accuracy], feed_dict = train_feed_dict)
            Train_accuracy += batch_accuracy
        Train_accuracy/=nbatch   # avg acc


        ###################### I modified it .. #########################
        # test_feed_dict = {self.image : self.test_data.image, self.label : self.test_data.label}  #
        for batch in tqdm(range(nbatch_t), ascii = True, desc = "batch"):
            start = self.batch_size*batch
            end = self.batch_size*(batch+1)
            test_feed_dict = {self.image: shuffle_test[start:end], self.label: shuffle_label_t[start:end]}
            test_accuracy = self.sess.run(self.accuracy, feed_dict=test_feed_dict)
            Test_accuracy += test_accuracy
        Test_accuracy/=nbatch_t   # avg test acc

        logger.info("Epoch({}/{}) train_accuracy : {}%, test_accuracy : {}%".format(epoch+1, self.epoch, Train_accuracy, Test_accuracy))
        if epoch%self.save_every == self.save_every-1:
            saver.save(self.sess, os.path.join(SAVE_DIR, 'model'), global_step = epoch+1)

hope it's not too late to you 。。。。

[zli2014]

thank u , it run well