I've followed the tutorial, but after the step of # Construct the variables for the NCE loss, which it's the block code of

with tf.device('/cpu:0'):
    # Loss function 
    num_sampled = 64        # Number of negative examples to sample. 
    loss = tf.reduce_mean(
        tf.nn.nce_loss(nce_weights, nce_biases, embed
                       , train_labels, num_sampled, vocabulary_size))
    # Optimizer
    optm = tf.train.GradientDescentOptimizer(1.0).minimize(loss)
    # Similarity measure (important)
    norm = tf.sqrt(tf.reduce_sum(tf.square(embeddings), 1, keep_dims=True))
    normalized_embeddings = embeddings / norm
    valid_embeddings = tf.nn.embedding_lookup(normalized_embeddings
                    , valid_dataset)
    siml = tf.matmul(valid_embeddings, normalized_embeddings
                    , transpose_b=True)
    
print ("Functions Ready")

Got the error of TypeError: Input 'b' of 'MatMul' Op has type float32 that does not match type int32 of argument 'a'.

In word2vec_basic, word2vec_simple

from tf.nn.nce_loss(nce_weights, nce_biases, embed, train_labels, num_sampled, vocabulary_size)) to tf.nn.nce_loss(nce_weights, nce_biases, train_labels, embed, num_sampled, vocabulary_size))

from summary_writer = tf.train.SummaryWriter('/tmp/tf_logs/word2vec', graph=sess.graph) to summary_writer = tf.summary.FileWriter('/tmp/tf_logs/word2vec', graph=sess.graph)

from tf.initialize_all_variables().run() to sess.run(tf.initialize_all_variables())

About https://github.com/sjchoi86/Tensorflow-101/blob/master/notebooks/mlp_mnist_simple.ipynb

Recently updated Tensorflow (1.4) will make the following error.

ValueError: Only call softmax_cross_entropy_with_logits with named arguments (labels=..., logits=..., ...)

Just simply adding named arguments will solve this issue.

cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits =pred, labels=y))

FYI, http://www.edwith.org/deeplearningchoi/lecture/15552/ has the right code.

I have done my own dataset and with the first step of training, i have the 100 percent accuracy for test and train data!! it was impossible I think, I search for the mistake in my code but I didn't find it! do you think where is my mistake?

Hi, @sjchoi86 ,

Thank you very much for releasing such a useful repos. As it's been quite a time since you first released this repos, so some conflict may occur when running this repos with current TF (0.12.1 or 1.0).

For example, when I run "semseg_basic.pynb" at the last cell, I came across this TF error

In order to run this repos, could you clarify the version of pre-requisite packages?

Tensorflow version: 0.xx? python version: 2.7/3.4/3.5 ? cuda version: 7.0/7.5/8.0? cudnn version ? anaconda: used or not?

THX!

I'm currently trying to generate a dataset based on your tutorial, but i was wondering how to deal with a huge data(10 gb of images). My laptop can't handle the huge amount of data(because the tutorial told us that we need to store the data into array variable first). Is there anyway to handle this? Thanks