Transcript:
All right, welcome to my new video on how to use. Python for deep learning, okay, So Pi Touch is just a numerical computation framework developed by Facebook Research and the main reason behind Pi Touch was to make computations very fast and very easy. Okay, so they decided to develop a torch with a concept of tensors. Okay, so Tensor’s is just a multi-dimensional arrays. Okay, so if you look at. Python Python has a library called numpy. Okay, Numpy is also used for numerical computations. They are linear algebra, which so, like adding two vectors matrices dividing matrices multiplying a matrix by a scalar dividing in finding the inverse of a matrix. Okay, so the reason why I thought came about Is Numpy is very slow and you can only run on Pi on Cpus. Okay, but with the help of Python, you can run pilotage on both CPU and GPU, okay, so Python can run on CPU and GPU, okay, but with Numpy Numpy can only run on CPU, so instead of slowing down your algorithm or your machine learning or deep learning code, you want to copy the tensors to a GPU in a distributed manner and then all the computation happens on GPU and then you, you collect back the results onto your CPU, okay, so the idea behind GPU, we can read more about it. It’s very faster than the CPU. Okay, so let’s go and search for. I say pile touch from Google, and then you’re gonna see this link by touch or click on it, so once you click on it. This is the their main interface getting started ecosystem, mobile blog, tutorials, Doc Documentation, Resource and Github. OK, so, you know, pilotage can run on? Alibaba Cloud, Amazon Web Services, Google Cloud and then Microsoft Azure. Okay, let me see. They also have a slot channel. If you want to join and ask any questions, you can do that. Let me see so if I click on getting started. You see this page low start? Locally stat, via cloud platform is previous Pi thought versions and then mobile. OK, so this is how you you install Pi touch locally on your machine, Okay, It could be Linux, Mac or Windows. So these are your operating system. OK, it a Linux, Mac or Windows, and then the built version. You have one point for stable and then preview nicely and then the package you can have Condor Peep, live touch and sauce. You can build fighters from sauce or just running Pip install or Fonda install, and then it supports this language is Python C++ or Java, and then Buddha, which its fighter supports CUDA version 9.2 and then 10.1 and none. Okay, so, CUDA. If you don’t know, it’s a said, it’s an engine that that is in between. It’s it’s a middle layer between Pi thoughts and GPU, so it connects I touch through the CUDA to the GPU. So who does sends all your Python code to the GPU? And then once the computation is done on the GPU, it sends the results but true CUDA to your Python interface. Okay, so if you want to use the Python, A GPU, you need to have CUDA installed by Nvi Di website. Okay, so I’m going to teach you all this and true either. My Linux box. Just say nano, or we can do it through my Macbook. OK, so if I click on Linux, and then I click on source and then I say C++ so once you. Do you select all these options? This is how you you you’re gonna install the Pi touch on your machine. So with this selection, it means that I have to install this library on Linux, But if I decide to select Mac, this is what from Cydia. You’re going to follow to install it, so I’m good or if you are on Windows, and then you want to do people install, and they want to do Python and then put a version 10.1 This is, this is going to be your selection. So all you need is to copy this link and then paste it in your terminal and then install. Pi touch, or if you want to do Mac. Do people install Python and CUDA version? All you need is people install touch and then tortured version, so you are installing to just torch and then torch vision. Okay, so that is how you will install phyto so. I’m going to copy this and then open my terminal. Okay, so I’m a gonna do this. So if I piss Pip install touch and then touch vision, make sure that you have Python 3 and above on your on your Macbook, Okay, so the version of Python we’re gonna use is Python 3 not Python 2 so if I hit enter, okay, so our requirements are satisfied because I already have it on my on my on my machine. So if I do Python, so you can see the I’m using Python 3 point seven point four. Okay, so if I do import if I do import, so if I do import touch, okay, starch is not imported or I can even do touch us. T and then T dot. So if I do double top, you can see these are all the methods and then fields that I have in a in Pi touch. Okay, so I can say cosine of I see 90 Okay, see, it gives me an arrow. So if I do t dot numb, let me see from Numpy, and then I put in this Irish comma, 4.15 You see, that also gives me an arrow, Okay, so if I do, it’s expecting an ende re, but it got a list, so what it means is that let me put N dimensional array double double brackets, and then that also gave me an arrow, so I’ll show you how to correct all these things. Once we start our first tutorial on how to [Music] declare arrays and then convert them from numpy to tenses. Okay, so I’m just playing around with all these arrows. I’ll show you why I made all these arrows. And then we can move forward in the tutorial, so thanks for watching. I don’t want to make it the video too long. So thanks for watching. I’m gonna make it short shot so that you can listen to it and enjoy the video so in our next video. I’m going to show you how we can use Pi Touch from the terminal and then from the Jupiter notebook and then also from the Pycharm IDE. So thanks for watching. See you in the next tutorial bye.