PyTorch Tensors Explained - Neural Network Programming

[@deeplizard]

Check out the corresponding blog and other resources for this video at: deeplizard.com/learn/video/jexkKugTg04

[@ratkush]

next post (wrapping up heading) link of deeplizard.com/learn/video/UWlFM0R_x6I redirecting to the same page.

[@Makwayne]

11pm and I'm binge watching this series instead of Netflix

[@deeplizard]

Haha I like it! 🤓💻🍿

[@deeplizard]

btw. What's good on Nextflix? Black Mirror? The 100?

[@Makwayne]

@@deeplizard the classic Top Gear ;)

[@manuelkarner8746]

@@deeplizard the 100 is pretty awsome, currently i am watching it the second time

[@shuangwang9886]

t1+t2 it works now!

[@joebender9052]

4:15 that is some intense music for typing :)

[@haneulkim4902]

Thought there was something wrong with my ear.... Had to unplug, plug my earbuds and looked around multiple times lol.

[@ericazheng9668]

Excellent explanation! It's super helpful.

[@deeplizard]

Thank you!

[@ShivamSUNDRAM-uj1xy]

the t1+t2 operation is now supported in pytorch, it will typecast every element to float and then add

[@mohammedhamdan4514]

{ "question": "Data preprocessing step are the first-line code when programming the neural network with PyTorch. So, the purpose of that code is to: ", "choices": [ "transform whatever the input data into tensors that can feed into NNs. ", "transform only numerical input data into tensors that can feed into NNs. ", "transform only textual input data into tensors that can feed into NNs. ", "transform only image input data into tensors that can feed into NNs. " ], "answer": "transform whatever the input data into tensors that can feed into NNs. ", "creator": "Hamdan", "creationDate": "2019-10-07T19:05:31.514Z" }

[@deeplizard]

Thank you Mohammed! The question is live. I did we word it to make the answers more concise.

[@pritamchakraborty6159]

very high-quality videos, and truly good explanations you give. These are very helpful videos. Thank you

[@Brahma2012]

Gradual development of conceptual understanding. It is going great..

[@vivekmittal3478]

Awesome content and also very well explained

[@ayushagarwal4247]

The output I'm getting is not returning an error while adding the tensors with different data types. I' getting a tensor with float data type. PS: I'm using version 1.3.1

[@deeplizard]

Hey ayush - Thank you for pointing this out. I'll be making a change to the text version of this episode on the website. PyTorch has recently changed this behavior. As of version 1.3.0, PyTorch has added Type Promotion. See the details here: github.com/pytorch/pytorch/releases Search for the text "Type Promotion".

[@saugatbhattarai327]

@@deeplizard Thank you for pointing out. Nice Tutorial series Thank you very much for making it free.

[@kinlam6304]

{ "question": "What is one of the criteria for performing tensor computation in pytorch?", "choices": [ "tensors must be the same type", "tensors must have the same shape", "tensors must have the same rank", "tensors mush have the same number of elements" ], "answer": "tensors must be the same type", "creator": "jlam", "creationDate": "2019-10-17T04:01:12.901Z" }

[@deeplizard]

Hey jlam - Thank you for contributing by making a question! 🙏 I've added the question to the quiz on the site with a slight rewording. You may need to clear your browser cache to see it! You are awesome!

[@SandwichMitGurke]

Im speeding through these videos with 1.5x and it's just so interesting :D

[@roros2512]

when I execute torch.Tensor(data) and all the variants, it doesn't show up the type of the data, what am I missing? thanks for the videos, it's a great work

[@willTryAgainTmrw]

I'm a bot, and I'm loving this series.

[@deeplizard]

🤖 bleep, bloop. I'm happy to hear that!

[@hosseinaboutalebi9998]

Great videos. One thing I would like to mention that it would be helpful if you also specify the pytorch version you are using. The syntax may change over the time.

[@deeplizard]

Thanks for the feedback, Hossein! Everything shown in the videos works with PyTorch v1.0.

[@haroldsu1696]

Thank you for the wonderful lectures,and it really helps.

[@fahadmuntasir2336]

PyTorch now allows Tensor operation between two different datatypes!

[@AshutoshRautela]

Amazing set of Tutorials

[@deeplizard]

Thank you Ashutosh! Really appreciate your feedback!

[@tymo3391]

awesome tutorial

[@karthickkuduva9819]

What would be best?. learning pytorch and tensorflow or being good at any one.

[@deeplizard]

Seek to transcend any single instance of a library. Extract the general pattern and take advantage of the synergistic effects.

[@karthickkuduva9819]

@@deeplizard thanks 😊

[@GeNiaaz]

Hi, I tried adding t1 and t2 of type int and float respectively like he showed in the example at 4:32, but when I "t1+t2", Jupyter Notebook didn't complain, I simply got tensor([2., 4., 6.]) as output. Can someone please explain?

[@deeplizard]

Hey Niaaz - This is working differently due to a PyTorch update. Please check the blog post where I added an update about this. See the section called "PyTorch Tensor Type Promotion". Link: deeplizard.com/learn/video/jexkKugTg04 You can also check the "Updates" section on each blog that will detail any changes that have occurred since the video was released. Hope this helps!

[@heller4196]

I can do this all day

[@MuhammadSalman-be1qz]

In newer versions adding two different dtype tensors does not give any error. It just add them give answer. But adding two tensors present on different devices(one on CPU and one on CUDA) is not possible it gives error syaing "Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu!"

[@Fighter_Believer_Achiever]

thank you so much

[@jayachandra677]

I did not get error while adding int64 and float32 tensors. pytorch might have added this capability. btw im using 1.6.0 version

[@deeplizard]

Hey Jay - You are right. PyTorch did add this capability. We have an "Updates" section for each video on the site. You can always check there to see if an update for any particular video has been published. This particular issue was already published. An explanation and details in the Updates section here: deeplizard.com/learn/video/jexkKugTg04

[@jayachandra677]

@@deeplizard thank you

[@tonihuhtiniemi1222]

In which lesson do we start running the model? :) Cant wait that! :P

[@deeplizard]

Hey Toni - Model details start on #17 in the series. This one: deeplizard.com/learn/video/k4jY9L8H89U

[@kalymdros]

I am following along and after trying t1 + t2 , I actually got : tensor([2.,4.,6.]) instead of Runtime Error. I can see the video is 5 years old, was there an update or something or did I do something wrong?

[@deeplizard]

Hello there! You are on the right track. Pytorch made an update in this regard. You can read about the update on the lesson page for this video here: deeplizard.com/learn/video/jexkKugTg04 See the section titled: PyTorch Tensor Type Promotion Happy learning! Chris

[@anthonyyaghi6342]

You deserve way more subscribers :/

[@psoma2744]

I have a doubt -- t1 = torch.tensor([1, 2, 3]) / t1 = torch.Tensor([1, 2, 3]) , I get different t1.dtype .. and i am able to add them .. ??

[@prathameshdinkar2966]

Now in 2022, you can add tensors with different dtypes also, like int + float!

[@zafarahmad6344]

Hi Sir! I can get the result of t1+t2, regardless that they have different data types t1= int64 while t2= float32 What ...I am doing wrong anything..? Any idea would be appreciated.

[@georgeigwegbe7258]

Well done!!!

[@deeplizard]

Hi George - Are you a bot or do you just watch on 4x?

[@georgeigwegbe7258]

🤣🤣..Nah, I am just a human on AI journey using PyTorch. Once again Well DONE!!! 💪

[@deeplizard]

Ah, so a human with AI superpowers! ✨ Lol thank you!

[@pradeepkumar-ew1ze]

In google colab, I could see the tensor's of different types gets added without error.

[@ai-video-stories]

yeah. it's a new feature

Pytorch 1.4 nightly tensor operations changed. no need to change int to float. it makes automaticly.

[@deeplizard]

Excellent.

[@louerleseigneur4532]

merci merci

[@りょりょりょ-b6s]

4:30 That's not true now. PyTorch will figure out the data types and do an implicit data type conversion. version: 1.5.0

[@utkar1]

I added t1 & t2 as shown got a result, is it a new addition to pytorch? Thanks for the awesome playlist `tensor([2., 4., 6.])` `torch.float32`

[@deeplizard]

It is new 😊 We've already updated the website for this one. You can see the details on the "Updates" section here: deeplizard.com/learn/video/jexkKugTg04 - Chris

[@supdenfoosukka]

I FUCKING UNDERSTAND WHAT THE SOFTWARE TEAM IS DOING

[@escapefelicity2913]

Get rid of the background noise

[@auggiewilliams3565]

UR channel is getting les recognitoin than it shud