Name: TensorFlowの内部：グラフの書き換え（関数ではなくマクロ (Inside TensorFlow: Graph rewriting (Macros, not functions))
Uploaded: 2021-01-14T10:03:59.000Z
Duration: 44 min 3 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：

様態の副詞

about how we do graph rewriting for functions.

So functions is a very important part of TF2.0.

It's very important for the end user, how user thinks

But that creates a bunch of problems for us at runtime.

And so basically, we have to rewrite all the graph

and all of the functions to make it executable

So this is, on the left, this button code.

So this kind of V1 Python that people usually write.

So you define your variables, some constants.

So if you don't add explicit control dependency,

So you have to think what are your dependencies

and have to make everything very explicit.

And so on the right, this is some textural GraphDef

So there's not enough space to go to the full GraphDef

like if you go to the back string into TensorFlow.

If you get to GraphDef to the back string,

you can get some short representation of the graph,

This is a single kernel that returns DT results data type.

Constants, they are just simple Const nodes with the values.

that takes results handle from VarHandleOp, and the constant,

And when you do it, it is another TensorFlow kernel

And because you have explicit control dependencies,

this series depends on update a and update c, both of them.

But explicitly, dividing control dependencies is very annoying.

And you have a TensorFlow notation and_and_get.

So I want to add two different constants, two variables, a

So the pattern looks something like on the left.

So you don't have to explicitly add control dependencies, which

And you don't have to think about program order.

And it has all the necessary control dependencies.

get the program order you would expect from [INAUDIBLE]..

So the FunctionDef, so the graph representation

And the GraphDef is basically a sequence of null Defs.

It's a way to group together some subgraph with input

So this function on the left from the pattern

That will be represented, something like this.

So we have two input parameters over source type a and b.

And the return type of the function is pair of ints.

We are assigning the new value to [INAUDIBLE] source.

from the previous operation, touching the same resource.

And this is added automatically by TensorFlow 2.0

In TensorFlow, they have two types of returns.

So you return two Tensors, read a and read c.

But also, it has a special kind of return notation, control

But you might group some of the ops inside your FunctionDef.

And you can specify that these ops have a name.

And when you have a control dependency on the function

that is part of control [INAUDIBLE] will run.

And the main graph will look something like this.

字幕リスト動画再生

TensorFlowの内部：グラフの書き換え（関数ではなくマクロ (Inside TensorFlow: Graph rewriting (Macros, not functions))

multiple

constant

basically

completely