Name: TensorFlowの内部: tf.Keras (パート2) (Inside TensorFlow: tf.Keras (part 2))
Uploaded: 2021-01-14T10:03:27.000Z
Duration: 51 min 34 s
Description: VoiceTubeの動画で発音を聞きながら英語表現を覚えよう！学べる英語：

程度の副詞

We talked about features that are specific to the functional

things like static input compatibility checks

a layer, whole-model saving, model plotting,

We talked about how masking works in the functional API

So for instance in this example, the Embedding layer is going

to be generating a mask here because you passed this

And in particular, to layers that consume the mask,

However, in this LSTM layer, because it's

going to be returning all the sequences, but only

and so this layer is not going to see a mask anymore.

Masking is pretty advanced, so most people who need it

And the idea is to enable them to benefit

from masking by just doing, like, hey, this Embedding

layer, I want the zeros to mean this is masked,

is going to magically know about this as long as they

you actually need to understand how it works.

writing a layer that consumes a mask, like an LSTM layer,

You just make sure you have this mask argument in the signature

that's going to match the structure of your inputs.

So it's because it's going to be a single tensor.

And the single tensor is going to be a Boolean tensor, where

you have one mask entry per timestep per sample.

If you have a layer that can safely pass through a mask,

for instance, a dense layer or in general any layer that

does not affect the time dimension of its inputs,

you can just enable your layer to pass through its mask

layers might be affecting the time dimension

of the inputs, in which case the meaning of the mask

And if you do have a layer that changes the time

dimension or otherwise a layer that creates a mask from input

values, it's going to need to implement this compute_mask

method that is going to receive the mask, receive the inputs.

If, for instance, you have an Embedding layer,

So it's going to be using the input values

If you have a concatenate layer, for instance,

it's not going to be looking at the input values,

And if one of the masks is [? non, ?] for instance,

we're going to have to generate a mask of 1's.

AUDIENCE: So just maybe a little bit more detail about masking--

so if you say supports_masking is true like

in the lower-left-hand corner, is that just using some default

FRANCOIS CHOLLET: Yes, and the default is pass through.

If you said this, this enables your layer

to use a default [INAUDIBLE] compute_mask, which

is like the first or second dimension gets masked?

gets masked if zero is the base dimension.

AUDIENCE: And then where does this mask argument come from?

it's not clear to me at all how this mask is being [INAUDIBLE]..

by the Embedding layer from the values of the integer inputs.

AUDIENCE: Right, so Embedding layers has a compute_mask

FRANCOIS CHOLLET: Yeah, which is exactly this one actually.

字幕リスト動画再生

TensorFlowの内部: tf.Keras (パート2) (Inside TensorFlow: tf.Keras (part 2))

essentially

pattern

basically

arbitrary