So this is specifically targeted at solving floating point equations.

所主要用来解决浮点方程的。

They are giant, inscrutable matrices of floating point numbers that we nudge in the direction of better performance until they inexplicably start working.

它们无法参透的大型浮点数矩阵，我们在朝着更好的性能试探着前进，直到它们莫名其妙地运行起来。

He did a great version of BASIC, but then never could buckle down and write the floating-point BASIC we needed, so we ended up later having to make a deal with Microsoft.

“他开发出了一个很棒的 BASIC 版本，但之后就没能认真编写我们需要的浮点 BASIC，最后我们不得不和微软做交易。

So what did we do? We're using the fact that, you know, we have a gigaflop sitting here.

所我们怎么做的呢？我们用到了千次浮点运算。

EFLOPS is a unit of the speed of computer systems, equal to one quintillion floating-point operations per second.

EFLOPS表示每秒执行一千万亿次的浮点运算。

So, you have learned that in Python programming language you have 2 types of numbers, integers which are whole numbers like 10, they don't have a decimal point, and floating point numbers or floats.

所，你已经了解到在 Python 编程语言中你有 2 种类型的数字，整数整数，如 10，它们没有小数点，及浮点数或浮点数。

Python doesn't know what to do with it. So in this case, we should convert this number by an integer or float and then multiply by 0.45.

Python不知道如何处理它。所在种情况下，我们应该转换个数字用整数或浮点数再乘0.45。

We need to divide the weight by 0.45. q So, weight divided by 0.45 and just to refresh your memory, this division operator returns a floating point number but if we use double slashes we'll get an integer.

我们需要将权重除 0.45。 q 所，权重除 0.45 为了刷新你的记忆，个除法运算符返回一个浮点数，但如果我们使用双斜线，我们将得到一个整数。

Plus, rounding errors are inherent in floating point computations.

此外，浮点计算中固有地存在舍入误差。

It's not a big deal if they have teraflops of whatever.

他们有多少万亿次浮点运算能力并不重要。

Measured in teraflops, a count of how many operations a chip can do in a second, one of NVIDIA's most capable chips is roughly as powerful as 300 or so top-end iPhones.

每秒万亿次浮点运算（teraflops）来衡量，即芯片每秒可执行的运算次数，NVIDIA 最强大的芯片之一的性能大致相当于 300 台高端 iPhone 的总和。

We have strings numbers and boolean Numbers can be integers floats or complex numbers Here's the second question. You have this variable fruit set to Apple. What do you think?

我们有字符串数字和布尔数字可整数浮点数或复数第二个问题。您将个变量 fruit 设置为 Apple。你怎么认为？

Can we think about using our AI analogs to reverse this curse of Eroom's law and jump on this exponential wave of GPU FLOPS powering our neural networks?

我们能否考虑使用AI类似技术来逆转尤姆定律的诅咒，并抓住GPU浮点运算能力推动我们的神经网络的指数级增长？

I think that something is happening with the floating point error because when you get to large factorials I think something weird happens out here.

我认为里出现了浮点误差问题，因为当阶乘变得很大时，里就会出现一些奇怪的情况。

This time I'm using floats, but let's go ahead and just run this.

次我用的浮点数，但让我们继续运行个。

We can only have one public field, native array of floats.

我们能有一个公共字段，即原生浮点数组。

It's going to become less efficient and flops is going to be less of a useful term.

效率将会降低，而浮点运算次数（flops）一指标的实用性也会随之减弱。

If we bypass their mainframe and direct route to the Homer cluster, we can clock this at around 600 teraflops.

如果我们绕过他们的主机直接进入分系统群，我们可把运算速度保持在600浮点运算次数每秒。

And so imagine porting floating point programs when the floating point didn't do the same thing.

想象一下将浮点程序移植到浮点运算行为不同的系统上的情景。

A good GPU can do teraflops today.

如今，优秀的 GPU 可实现万亿次浮点运算。