Name: TensorFlowの内部：熱心な実行時間 (Inside TensorFlow: Eager execution runtime)
Uploaded: 2021-01-14T10:34:59.000Z
Duration: 45 min 11 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：

ALEX PASSOS: Hi, my name is Alex, and I'm here again,

this time to talk about the TensorFlow eager execution

are lots and lots of things we could cover,

so I'm going to lightly graze many, many different parts

that you can use to familiarize yourself with something,

but if you're in the room right now, by all means,

Like, this is-- there's some buffer time at the end

And I think the more we can maximize shared understanding

So the way I thought we could go about this

actually happens in TensorFlow, starting from TF2--

when you type a very simple line of code, in this case

And I think if you were to start doing this, probably

the first thing you'd do is graph the TensorFlow code

that we have some definitions of ReLU in Keras,

but you won't find a single definition of ReLU itself

let's find out what actually happens when we run ReLU

business, but the way ReLU comes from is because it's enough

that we've implemented in C++ and we didn't need to put

We can just generate the Python code to call ReLU.

we use to register all the ops in TensorFlow.

So for every core operation in TensorFlow

have a registration that looks like this--

It takes a name and you say how many inputs, how many outputs,

If you want to know more about attributes-- what things

we can make our ops polymorphic and have the same operation

have different types of outputs, so different numbers of outputs

There is a lot of documentation about this in TensorFlow.org,

if you search for how to define a new op.

is that we also register a shape_inference function.

ops we have-- it just has one input, one output,

So we can use a pre-built shape_inference function

that just says the shape does not change.

Other ops will have vastly more complicated

run these functions offline for our building graphs

without actually having the values of any tensors

about the shapes of intermediate tenses and outputs

This is maybe the best tool we have for catching bugs now.

So if you want to look at the shape_inference code,

we run some complicated code that generates Python code

And if you look in bazel-genfiles, files,

you will find a file named gen_nn_ops.py and this file has

and there's a lot of stuff going in there.

so that we can define ReLU not just for normal tensors,

and ragged tensors and other composite types.

The second line has tf_export and what this does

Every symbol that you get when you are using TensorFlow by tf

from a tf_export decorator like this one.

There will be a future video on how exactly this works

But you can probably guess that it's because TensorFlow

and this generated code for ReLU has a bunch of cases in it.

You have an eager fast path, an eager slow path,

you have a graph mode path, and kind of a side hook

But here, let's focus on the eager paths.

字幕リスト動画再生

TensorFlowの内部：熱心な実行時間 (Inside TensorFlow: Eager execution runtime)

bunch

context

multiple

tricky