Discrete Illustration Studying with VQ-VAE and TensorFlow Chance

About two weeks in the past, we launched TensorFlow Chance (TFP), displaying the best way to create and pattern from distributions and put them to make use of in a Variational Autoencoder (VAE) that learns its prior. In the present day, we transfer on to a distinct specimen within the VAE mannequin zoo: the Vector Quantised Variational Autoencoder (VQ-VAE) described in Neural Discrete Illustration Studying (Oord, Vineals, and Kavokcuglu 2017). This mannequin differs from most VAEs in that its approximate posterior just isn’t steady, however discrete – therefore the “quantised” within the article’s title. We’ll rapidly take a look at what this implies, after which dive instantly into the code, combining Keras layers, keen execution, and TFP.

Many phenomena are greatest considered, and modeled, as discrete. This holds for phonemes and lexemes in language, higher-level constructions in photographs (assume objects as an alternative of pixels),and duties that necessitate reasoning and planning.
The latent code utilized in most VAEs, nevertheless, is steady – normally it’s a multivariate Gaussian. Steady-space VAEs have been discovered very profitable in reconstructing their enter, however typically they undergo from one thing known as posterior collapse: The decoder is so highly effective that it might create lifelike output given simply any enter. This implies there isn’t a incentive to be taught an expressive latent house.

In VQ-VAE, nevertheless, every enter pattern will get mapped deterministically to certainly one of a set of embedding vectors. Collectively, these embedding vectors represent the prior for the latent house.
As such, an embedding vector comprises much more info than a imply and a variance, and thus, is way tougher to disregard by the decoder.

The query then is: The place is that magical hat, for us to drag out significant embeddings?

From the above conceptual description, we now have two inquiries to reply. First, by what mechanism will we assign enter samples (that went by means of the encoder) to acceptable embedding vectors?
And second: How can we be taught embedding vectors that really are helpful representations – that when fed to a decoder, will lead to entities perceived as belonging to the identical species?

As regards task, a tensor emitted from the encoder is solely mapped to its nearest neighbor in embedding house, utilizing Euclidean distance. The embedding vectors are then up to date utilizing exponential transferring averages. As we’ll see quickly, which means they’re truly not being realized utilizing gradient descent – a function value declaring as we don’t come throughout it day by day in deep studying.

Concretely, how then ought to the loss perform and coaching course of look? It will in all probability best be seen in code.

The whole code for this instance, together with utilities for mannequin saving and picture visualization, is offered on github as a part of the Keras examples. Order of presentation right here could differ from precise execution order for expository functions, so please to really run the code think about making use of the instance on github.

As in all our prior posts on VAEs, we use keen execution, which presupposes the TensorFlow implementation of Keras.

As in our earlier put up on doing VAE with TFP, we’ll use Kuzushiji-MNIST(Clanuwat et al. 2018) as enter.
Now could be the time to have a look at what we ended up producing that point and place your guess: How will that evaluate towards the discrete latent house of VQ-VAE?

np <- import("numpy")
 
kuzushiji <- np$load("kmnist-train-imgs.npz")
kuzushiji <- kuzushiji$get("arr_0")

train_images <- kuzushiji %>%
  k_expand_dims() %>%
  k_cast(dtype = "float32")

train_images <- train_images %>% `/`(255)

buffer_size <- 60000
batch_size <- 64
num_examples_to_generate <- batch_size

batches_per_epoch <- buffer_size / batch_size

train_dataset <- tensor_slices_dataset(train_images) %>%
  dataset_shuffle(buffer_size) %>%
  dataset_batch(batch_size, drop_remainder = TRUE)

Hyperparameters

Along with the “standard” hyperparameters we’ve got in deep studying, the VQ-VAE infrastructure introduces just a few model-specific ones. To begin with, the embedding house is of dimensionality variety of embedding vectors occasions embedding vector measurement:

# variety of embedding vectors
num_codes <- 64L
# dimensionality of the embedding vectors
code_size <- 16L

The latent house in our instance shall be of measurement one, that’s, we’ve got a single embedding vector representing the latent code for every enter pattern. This shall be wonderful for our dataset, but it surely needs to be famous that van den Oord et al. used far higher-dimensional latent areas on e.g. ImageNet and Cifar-10.

Encoder mannequin

The encoder makes use of convolutional layers to extract picture options. Its output is a three-D tensor of form batchsize * 1 * code_size.

activation <- "elu"
# modularizing the code just a bit bit
default_conv <- set_defaults(layer_conv_2d, record(padding = "similar", activation = activation))

base_depth <- 32

encoder_model <- perform(title = NULL,
                          code_size) {
  
  keras_model_custom(title = title, perform(self) {
    
    self$conv1 <- default_conv(filters = base_depth, kernel_size = 5)
    self$conv2 <- default_conv(filters = base_depth, kernel_size = 5, strides = 2)
    self$conv3 <- default_conv(filters = 2 * base_depth, kernel_size = 5)
    self$conv4 <- default_conv(filters = 2 * base_depth, kernel_size = 5, strides = 2)
    self$conv5 <- default_conv(filters = 4 * latent_size, kernel_size = 7, padding = "legitimate")
    self$flatten <- layer_flatten()
    self$dense <- layer_dense(items = latent_size * code_size)
    self$reshape <- layer_reshape(target_shape = c(latent_size, code_size))
    
    perform (x, masks = NULL) {
      x %>% 
        # output form:  7 28 28 32 
        self$conv1() %>% 
        # output form:  7 14 14 32 
        self$conv2() %>% 
        # output form:  7 14 14 64 
        self$conv3() %>% 
        # output form:  7 7 7 64 
        self$conv4() %>% 
        # output form:  7 1 1 4 
        self$conv5() %>% 
        # output form:  7 4 
        self$flatten() %>% 
        # output form:  7 16 
        self$dense() %>% 
        # output form:  7 1 16
        self$reshape()
    }
  })
}

As at all times, let’s make use of the truth that we’re utilizing keen execution, and see just a few instance outputs.

iter <- make_iterator_one_shot(train_dataset)
batch <-  iterator_get_next(iter)

encoder <- encoder_model(code_size = code_size)
encoded  <- encoder(batch)
encoded

tf.Tensor(
((( 0.00516277 -0.00746826  0.0268365  ... -0.012577   -0.07752544
   -0.02947626))
...

 ((-0.04757921 -0.07282603 -0.06814402 ... -0.10861694 -0.01237121
    0.11455103))), form=(64, 1, 16), dtype=float32)

Now, every of those 16d vectors must be mapped to the embedding vector it’s closest to. This mapping is taken care of by one other mannequin: vector_quantizer.

Vector quantizer mannequin

That is how we’ll instantiate the vector quantizer:

vector_quantizer <- vector_quantizer_model(num_codes = num_codes, code_size = code_size)

This mannequin serves two functions: First, it acts as a retailer for the embedding vectors. Second, it matches encoder output to out there embeddings.

Right here, the present state of embeddings is saved in codebook. ema_means and ema_count are for bookkeeping functions solely (word how they’re set to be non-trainable). We’ll see them in use shortly.

vector_quantizer_model <- perform(title = NULL, num_codes, code_size) {
  
    keras_model_custom(title = title, perform(self) {
      
      self$num_codes <- num_codes
      self$code_size <- code_size
      self$codebook <- tf$get_variable(
        "codebook",
        form = c(num_codes, code_size), 
        dtype = tf$float32
        )
      self$ema_count <- tf$get_variable(
        title = "ema_count", form = c(num_codes),
        initializer = tf$constant_initializer(0),
        trainable = FALSE
        )
      self$ema_means = tf$get_variable(
        title = "ema_means",
        initializer = self$codebook$initialized_value(),
        trainable = FALSE
        )
      
      perform (x, masks = NULL) { 
        
        # to be stuffed in shortly ...
        
      }
    })
}

Along with the precise embeddings, in its name technique vector_quantizer holds the task logic.
First, we compute the Euclidean distance of every encoding to the vectors within the codebook (tf$norm).
We assign every encoding to the closest as by that distance embedding (tf$argmin) and one-hot-encode the assignments (tf$one_hot). Lastly, we isolate the corresponding vector by masking out all others and summing up what’s left over (multiplication adopted by tf$reduce_sum).

Relating to the axis argument used with many TensorFlow capabilities, please consider that in distinction to their k_* siblings, uncooked TensorFlow (tf$*) capabilities count on axis numbering to be 0-based. We even have so as to add the L’s after the numbers to evolve to TensorFlow’s datatype necessities.

vector_quantizer_model <- perform(title = NULL, num_codes, code_size) {
  
    keras_model_custom(title = title, perform(self) {
      
      # right here we've got the above occasion fields
      
      perform (x, masks = NULL) {
    
        # form: bs * 1 * num_codes
         distances <- tf$norm(
          tf$expand_dims(x, axis = 2L) -
            tf$reshape(self$codebook, 
                       c(1L, 1L, self$num_codes, self$code_size)),
                       axis = 3L 
        )
        
        # bs * 1
        assignments <- tf$argmin(distances, axis = 2L)
        
        # bs * 1 * num_codes
        one_hot_assignments <- tf$one_hot(assignments, depth = self$num_codes)
        
        # bs * 1 * code_size
        nearest_codebook_entries <- tf$reduce_sum(
          tf$expand_dims(
            one_hot_assignments, -1L) * 
            tf$reshape(self$codebook, c(1L, 1L, self$num_codes, self$code_size)),
                       axis = 2L 
                       )
        record(nearest_codebook_entries, one_hot_assignments)
      }
    })
  }

Now that we’ve seen how the codes are saved, let’s add performance for updating them.
As we stated above, they aren’t realized by way of gradient descent. As an alternative, they’re exponential transferring averages, frequently up to date by no matter new “class member” they get assigned.

So here’s a perform update_ema that can deal with this.

update_ema makes use of TensorFlow moving_averages to

• first, hold observe of the variety of at the moment assigned samples per code (updated_ema_count), and
• second, compute and assign the present exponential transferring common (updated_ema_means).

moving_averages <- tf$python$coaching$moving_averages

# decay to make use of in computing exponential transferring common
decay <- 0.99

update_ema <- perform(
  vector_quantizer,
  one_hot_assignments,
  codes,
  decay) {
 
  updated_ema_count <- moving_averages$assign_moving_average(
    vector_quantizer$ema_count,
    tf$reduce_sum(one_hot_assignments, axis = c(0L, 1L)),
    decay,
    zero_debias = FALSE
  )

  updated_ema_means <- moving_averages$assign_moving_average(
    vector_quantizer$ema_means,
    # selects all assigned values (masking out the others) and sums them up over the batch
    # (shall be divided by depend later, so we get a mean)
    tf$reduce_sum(
      tf$expand_dims(codes, 2L) *
        tf$expand_dims(one_hot_assignments, 3L), axis = c(0L, 1L)),
    decay,
    zero_debias = FALSE
  )

  updated_ema_count <- updated_ema_count + 1e-5
  updated_ema_means <-  updated_ema_means / tf$expand_dims(updated_ema_count, axis = -1L)
  
  tf$assign(vector_quantizer$codebook, updated_ema_means)
}

Earlier than we take a look at the coaching loop, let’s rapidly full the scene including within the final actor, the decoder.

Decoder mannequin

The decoder is fairly customary, performing a collection of deconvolutions and at last, returning a likelihood for every picture pixel.

default_deconv <- set_defaults(
  layer_conv_2d_transpose,
  record(padding = "similar", activation = activation)
)

decoder_model <- perform(title = NULL,
                          input_size,
                          output_shape) {
  
  keras_model_custom(title = title, perform(self) {
    
    self$reshape1 <- layer_reshape(target_shape = c(1, 1, input_size))
    self$deconv1 <-
      default_deconv(
        filters = 2 * base_depth,
        kernel_size = 7,
        padding = "legitimate"
      )
    self$deconv2 <-
      default_deconv(filters = 2 * base_depth, kernel_size = 5)
    self$deconv3 <-
      default_deconv(
        filters = 2 * base_depth,
        kernel_size = 5,
        strides = 2
      )
    self$deconv4 <-
      default_deconv(filters = base_depth, kernel_size = 5)
    self$deconv5 <-
      default_deconv(filters = base_depth,
                     kernel_size = 5,
                     strides = 2)
    self$deconv6 <-
      default_deconv(filters = base_depth, kernel_size = 5)
    self$conv1 <-
      default_conv(filters = output_shape(3),
                   kernel_size = 5,
                   activation = "linear")
    
    perform (x, masks = NULL) {
      
      x <- x %>%
        # output form:  7 1 1 16
        self$reshape1() %>%
        # output form:  7 7 7 64
        self$deconv1() %>%
        # output form:  7 7 7 64
        self$deconv2() %>%
        # output form:  7 14 14 64
        self$deconv3() %>%
        # output form:  7 14 14 32
        self$deconv4() %>%
        # output form:  7 28 28 32
        self$deconv5() %>%
        # output form:  7 28 28 32
        self$deconv6() %>%
        # output form:  7 28 28 1
        self$conv1()
      
      tfd$Unbiased(tfd$Bernoulli(logits = x),
                      reinterpreted_batch_ndims = size(output_shape))
    }
  })
}

input_shape <- c(28, 28, 1)
decoder <- decoder_model(input_size = latent_size * code_size,
                         output_shape = input_shape)

Now we’re prepared to coach. One factor we haven’t actually talked about but is the fee perform: Given the variations in structure (in comparison with customary VAEs), will the losses nonetheless look as anticipated (the same old add-up of reconstruction loss and KL divergence)?
We’ll see that in a second.

Coaching loop

Right here’s the optimizer we’ll use. Losses shall be calculated inline.

optimizer <- tf$practice$AdamOptimizer(learning_rate = learning_rate)

The coaching loop, as standard, is a loop over epochs, the place every iteration is a loop over batches obtained from the dataset.
For every batch, we’ve got a ahead cross, recorded by a gradientTapebased mostly on which we calculate the loss.
The tape will then decide the gradients of all trainable weights all through the mannequin, and the optimizer will use these gradients to replace the weights.

To this point, all of this conforms to a scheme we’ve oftentimes seen earlier than. One level to notice although: On this similar loop, we additionally name update_ema to recalculate the transferring averages, as these are usually not operated on throughout backprop.
Right here is the important performance:

num_epochs <- 20

for (epoch in seq_len(num_epochs)) {
  
  iter <- make_iterator_one_shot(train_dataset)
  
  until_out_of_range({
    
    x <-  iterator_get_next(iter)
    with(tf$GradientTape(persistent = TRUE) %as% tape, {
      
      # do ahead cross
      # calculate losses
      
    })
    
    encoder_gradients <- tape$gradient(loss, encoder$variables)
    decoder_gradients <- tape$gradient(loss, decoder$variables)
    
    optimizer$apply_gradients(purrr::transpose(record(
      encoder_gradients, encoder$variables
    )),
    global_step = tf$practice$get_or_create_global_step())
    
    optimizer$apply_gradients(purrr::transpose(record(
      decoder_gradients, decoder$variables
    )),
    global_step = tf$practice$get_or_create_global_step())
    
    update_ema(vector_quantizer,
               one_hot_assignments,
               codes,
               decay)

    # periodically show some generated photographs
    # see code on github 
    # visualize_images("kuzushiji", epoch, reconstructed_images, random_images)
  })
}

Now, for the precise motion. Contained in the context of the gradient tape, we first decide which encoded enter pattern will get assigned to which embedding vector.

codes <- encoder(x)
c(nearest_codebook_entries, one_hot_assignments) %<-% vector_quantizer(codes)

Now, for this task operation there isn’t a gradient. As an alternative what we are able to do is cross the gradients from decoder enter straight by means of to encoder output.
Right here tf$stop_gradient exempts nearest_codebook_entries from the chain of gradients, so encoder and decoder are linked by codes:

codes_straight_through <- codes + tf$stop_gradient(nearest_codebook_entries - codes)
decoder_distribution <- decoder(codes_straight_through)

In sum, backprop will deal with the decoder’s in addition to the encoder’s weights, whereas the latent embeddings are up to date utilizing transferring averages, as we’ve seen already.

Now we’re able to deal with the losses. There are three elements:

• First, the reconstruction loss, which is simply the log likelihood of the particular enter below the distribution realized by the decoder.

reconstruction_loss <- -tf$reduce_mean(decoder_distribution$log_prob(x))

• Second, we’ve got the dedication lossoutlined because the imply squared deviation of the encoded enter samples from the closest neighbors they’ve been assigned to: We would like the community to “commit” to a concise set of latent codes!

commitment_loss <- tf$reduce_mean(tf$sq.(codes - tf$stop_gradient(nearest_codebook_entries)))

• Lastly, we’ve got the same old KL diverge to a previous. As, a priori, all assignments are equally possible, this element of the loss is fixed and may oftentimes be allotted of. We’re including it right here primarily for illustrative functions.

prior_dist <- tfd$Multinomial(
  total_count = 1,
  logits = tf$zeros(c(latent_size, num_codes))
  )
prior_loss <- -tf$reduce_mean(
  tf$reduce_sum(prior_dist$log_prob(one_hot_assignments), 1L)
  )

Summing up all three elements, we arrive on the total loss:

beta <- 0.25
loss <- reconstruction_loss + beta * commitment_loss + prior_loss

Earlier than we take a look at the outcomes, let’s see what occurs inside gradientTape at a single look:

with(tf$GradientTape(persistent = TRUE) %as% tape, {
      
  codes <- encoder(x)
  c(nearest_codebook_entries, one_hot_assignments) %<-% vector_quantizer(codes)
  codes_straight_through <- codes + tf$stop_gradient(nearest_codebook_entries - codes)
  decoder_distribution <- decoder(codes_straight_through)
      
  reconstruction_loss <- -tf$reduce_mean(decoder_distribution$log_prob(x))
  commitment_loss <- tf$reduce_mean(tf$sq.(codes - tf$stop_gradient(nearest_codebook_entries)))
  prior_dist <- tfd$Multinomial(
    total_count = 1,
    logits = tf$zeros(c(latent_size, num_codes))
  )
  prior_loss <- -tf$reduce_mean(tf$reduce_sum(prior_dist$log_prob(one_hot_assignments), 1L))
  
  loss <- reconstruction_loss + beta * commitment_loss + prior_loss
})

Outcomes

And right here we go. This time, we are able to’t have the second “morphing view” one typically likes to show with VAEs (there simply isn’t any second latent house). As an alternative, the 2 photographs beneath are (1) letters generated from random enter and (2) reconstructed precise letters, every saved after coaching for 9 epochs.

Two issues bounce to the attention: First, the generated letters are considerably sharper than their continuous-prior counterparts (from the earlier put up). And second, would you will have been in a position to inform the random picture from the reconstruction picture?

At this level, we’ve hopefully satisfied you of the facility and effectiveness of this discrete-latents strategy.
Nevertheless, you may secretly have hoped we’d apply this to extra complicated information, comparable to the weather of speech we talked about within the introduction, or higher-resolution photographs as present in ImageNet.

The reality is that there’s a steady tradeoff between the variety of new and thrilling strategies we are able to present, and the time we are able to spend on iterations to efficiently apply these strategies to complicated datasets. In the long run it’s you, our readers, who will put these strategies to significant use on related, actual world information.