I hope you enjoy the course :) And check out Tabnine, the FREE AI-powered code completion tool that helps you to code faster: www.tabnine.com/?.com&PythonEngineer * ---------------------------------------------------------------------------------------------------------- * This is a sponsored link. You will not have any additional costs, instead you will support me and my project. Thank you so much for the support! 🙏
at 37:00 I found after adding 2 that not all members of the tensor had exactly x+2. I tried this several times with always one of the parts of the tensor had less than x+2. Then at 37:16 you also had an anomaly. Why is this?
Incredible tutorial, thank you! Some corrections: - 1:12:02 correct gradient function in the manual gradient calculation should be `np.dot(2*x, y_predicted - y) / len(x)`, because np.dot results in a scalar and mean() has no effect of calculating the mean. (TY @Arman Seyed-Ahmadi) - 1:23:52 the optimizer is applying the gradient exactly like we do, there is no difference. The reason the PyTorch model has different predictions is because 1) you use a model with a bias, 2) the values are initialized randomly. To turn off the bias use `bias=False` in the model construction. To initialize the weight to zero use a `with torch.no_grad()` block and set `model.weight[0,0] = 0`. Then all versions result in the exact same model with the exact same predictions (as expected).
Thanks for this second comment! To add to this: nn.Linear wants to solve y = wx + b here. This 'b' is the bias, and by setting bias = False, instead it learns y = wx as we want it to. This also means that model.parameters() will yield only [w] and not [w, b] anymore, so do not forget to change that in line 52 in the video as well.
This is a fantastic tutorial, thank you for sharing this great material! There is one mistake though that needs clarification: ========================================== At 1:12:02 it is mentioned that the code with automatic differentiation does not converge as fast because "back-propagation is not as exact as the numerical gradient". This is incorrect: the reason why the convergence of the two codes are different is because there is a mistake in the gradient() function. When the dot product np.dot(2x, y_pred_y) is performed, the result is a scalar and .mean() does not do anything. Instead of doing .mean(), np.dot(2x, y_pred_y) should simply be divided by len(x) to give the correct mean gradient. After doing this, both methods give the exact same convergence history and final results.
Thanks for the course Patrick! It was a great refresher! BTW, at 3:42:02, in the newer versions instead of pretrained=True it is changed to weights=True.
This is one of the very few videos which is teaching Pytorch from the ground up! Beautiful work, @Python Engineer. Highly recommend it for any newbie + refresher.
Update: Note a subtle detail, if in with torch.no_grad() you use w = instead of w -= a new w variable will be created with requires_grad = False, which is fixed by w.requires_grad = True Original: Using pytorch 1.11, and go figure @1:11 w.grad.zero_() errors, instead I had to put w.requires_grad = True
For the feedforward part, you need to send the model to the GPU when instantiating it: model = NeuralNet(input_size, hidden_size, num_classes).to(device) if your device is 'cuda' and you forget the '.to(device)' you will get an error.
at 1:01:41 he uses np.dot and when it should be np.multiply, that will make it consistent with the pytorch implementation. By doing np.dot, the items are multiplied and summed leaving just one value to which the mean function is applied, so the reason the numpy version get to 0 loss quicker is the gradient is not being averaged correctly.
I just completed the course on ML from scratch from Python Engineer. It was a great course for someone who learned all those algorithms in the past and wants to see how they get implemented using basic python lib and numpy.
On 4:14:00, I think you should use the ground truth as the labels rather than the predicted (line 130). Because the PR curve use the ground truth and predicted score to paint
Thanks a lot for the low level explanations. At 1:01:47 when you dot product the array turns into a single scalar. So mean() returns that number(the sum), not average. When you fix it you get the exact same results as with pytorch's implementation in 1:12:00
@@phi6934 I don't remember the details right now, but just dividing the expression with the size of the tensor must do the work. In the expression put smt like .../len(x) instead of .mean()
Thanks for the awesome course! The material is extremely well curated, every minute is pure gold. I particularly liked the fact that for each subject there is a smooth transition from numpy to torch. It's perfect for someone who wants a quick and thorough deeplearning recap and get comfortable with hands-on pytorch coding.
Dear with apologies kindly notice, At timestamp 1:12:05 make a correction in stating, that the backprop grad was not correct, Actually the numerical one was not correct. Because np.dot is computing a single number and then taking mean is the same number, instead use 2*x/4 in np.dot(2*x,(Y_pred-Y).mean()) to correct your numerical gradient. Using np.dot(2*x/4,(Y_pred-Y)) will produce same result as back propagated result. Mean will be usefull when W and X are matrices. Thank you
This is the best course on this topic I've seen so far. It is perfect when you want to understand what you're doing and the way things are brought is very pedagogic.
If z is a scalar then z.backward() is defined (and I understand the computation), while if z is not a scalar then z.backward() is not defined unless you provide appropriate inputs. However, it was not entirely clear to me what computation is occurring when we do z.backward(x) for example (where x is appropriate). This subject matter is around 33:00.
What is happening is that PyTorch is assuming that you have provided the intermediate gradients i.e. (dLoss/dz), then using these intermediate gradients PyTorch is able to compute the gradients further downstream and backward step is successful.
In the Gradient Descent and Training Pipeline sections, the presenter glosses over why it takes 5x more training steps to converge. There are a couple factors: - Autograd is less aggressive than the manual gradient calculation, effectively lowering the learning rate (you can go all the way up to 0.1 after you move to torch and autograd) - nn.Linear() includes a bias by default and a non-zero initialization of the weights, making it not a direct comparison. You can get much closer by adding `bias=False` to the model initialization and by zeroing out the weigth with `model.weight.data.fill_(0.0`
Someone has probably mentioned this already, but on line 23 at 1:04:08 .mean() is not doing anything since taking the dot product already returned a scalar. This is just dividing by one. Instead, you should be dividing by len(x) or len(y), or there may be another more efficient way to get the same result.
1:12:09 It's because the gradient in your formula is not correct, not because pytorch's backpropogation calculation. You should put the ".mean()" into the brackets of "np.dot()".
Your course is great! Congratulations! I just had to do a small correction in your code in part "13. Feed Forward Net" so that I could run it on GPU. It was necessary to add the "device" (that was preciously declared) as an argument in the nn.Linear function. Without this detail it is not possible to run the code in GPU. class NeuralNet(nn.Module): def __init__(self, input_size, hidden_size, n_classes, device): super(NeuralNet,self).__init__() self.l1 = nn.Linear(input_size, hidden_size, device=device) self.relu = nn.ReLU() self.l2 = nn.Linear(hidden_size, n_classes, device=device) def forward(self, x): out = self.l1(x) out = self.relu(out) out = self.l2(out) return out
How is it that for a feed forward neural network we zero the gradients first before computing gradients and updating weights @3:08:35, whereas in the case of linear/logistic regression, we zero the gradients after computing them and updating the weights @1:36:19 @1:52:41. Intuitively, this should not make any difference, but i wanted to confirm if that truely is the case. Is this just a nomenclature thingy?
Nice tutorial ! @1:11:40 at line # 37. Instead of using "w -= learning_rate * w.grad" , I used expanded form "w = w - learning_rate * w.grad" and thought it would be same. But in this case 'w.grad' return 'None'. w.require_grad is False and hence error. Though "w -= learning_rate * w.grad" is same as "w.data = w.data - learning_rate * w.grad". It seems torch Tensor ( with require_grad True) have some overridden "__iadd__" implementation.
Note at 2:08: `dataiter.next()` is no throwing an AttributeError: '_MultiProcessingDataLoaderIter' object has no attribute 'next'... I changed that line to `data = next(dataiter)`
This is an error I have found Time: 1:01:55 According to the equation,we actually need to find 1/N ,where N represents the number of term(here 4).According to the code,we are computing mean after converting the rest of the code to a dot product,which contains just a value.So instead of dividing with the desired value(4),we are dividing with 1.
At 1:01:55 you are taking the mean of a scalar, which doesn't do anything. Since you have 4 data points only this effectively means that your learning_rate was multiplied by 4. This is the reason why it seems to work better than PyTorch: this particular case is so well behaved that to speed up is sufficient to take larger steps.
This tutorial is supppppppppppper great! The best deep learning tutorial I've ever watched. Thank you so much. I enjoined the tutorial that I didn't want it to stop! I look forward to seeing more great videos like this from this channel
Thanks for this incredible resource. FYI I believe the gradient function computed at 1:01:38 is incorrect. I'm pretty sure it should be: def gradient(x, y, y_predicted): return ((y_predicted-y)*2*x).mean()
a probable mistake: Leaky ReLU isn't used for solving the problem of vanishing gradient problem but Dead Neurons problem. Which can happen when you use ReLU activation functions.
Absolutely great. But what was missing for me was how then to use a trained model. Conspicuous in its absence was how at the end to feed data into a trained model and get the answer it was trained to give. Is there another video that explains this?
1:42:10 Is there a reason you used `X, y` instead of `X, Y`? I believe it should `Y` as we're dealing with a tensor of dependent variables right? It would be `y` if we were dealing with a scalar though
I've taken a graduate course in deep learning and neural, and have watched other tutorials here and there, but this is by far the most helpful one. Granted, all the previous materials have probably contributed, but the way you teach is unparalleled!
![Build Your First Pytorch Model In Minutes! [Tutorial + Code]](http://i.ytimg.com/vi/tHL5STNJKag/mqdefault.jpg)
![Build Your First Pytorch Model In Minutes! [Tutorial + Code]](/img/tr.png)
I hope you enjoy the course :)
And check out Tabnine, the FREE AI-powered code completion tool that helps you to code faster: www.tabnine.com/?.com&PythonEngineer *
----------------------------------------------------------------------------------------------------------
* This is a sponsored link. You will not have any additional costs, instead you will support me and my project. Thank you so much for the support! 🙏
at 37:00 I found after adding 2 that not all members of the tensor had exactly x+2. I tried this several times with always one of the parts of the tensor had less than x+2. Then at 37:16 you also had an anomaly. Why is this?
Thank you very much. You did a great work!
.👆Never love anyone who treats you like you’re ordinary.
great video! thank you but please don't delete each line that you code! wait till the subject is finished then delete them once
I’m really enjoying it mate. Hope you are doing well. 🎉
Incredible tutorial, thank you! Some corrections:
- 1:12:02 correct gradient function in the manual gradient calculation should be `np.dot(2*x, y_predicted - y) / len(x)`, because np.dot results in a scalar and mean() has no effect of calculating the mean. (TY @Arman Seyed-Ahmadi)
- 1:23:52 the optimizer is applying the gradient exactly like we do, there is no difference. The reason the PyTorch model has different predictions is because 1) you use a model with a bias, 2) the values are initialized randomly. To turn off the bias use `bias=False` in the model construction. To initialize the weight to zero use a `with torch.no_grad()` block and set `model.weight[0,0] = 0`. Then all versions result in the exact same model with the exact same predictions (as expected).
Thanks for this second comment! To add to this: nn.Linear wants to solve y = wx + b here. This 'b' is the bias, and by setting bias = False, instead it learns y = wx as we want it to. This also means that model.parameters() will yield only [w] and not [w, b] anymore, so do not forget to change that in line 52 in the video as well.
This is a fantastic tutorial, thank you for sharing this great material!
There is one mistake though that needs clarification:
==========================================
At 1:12:02 it is mentioned that the code with automatic differentiation does not converge as fast because "back-propagation is not as exact as the numerical gradient". This is incorrect: the reason why the convergence of the two codes are different is because there is a mistake in the gradient() function. When the dot product np.dot(2x, y_pred_y) is performed, the result is a scalar and .mean() does not do anything. Instead of doing .mean(), np.dot(2x, y_pred_y) should simply be divided by len(x) to give the correct mean gradient. After doing this, both methods give the exact same convergence history and final results.
I wishhhh saw your comment earlier. I was just going crazy that what am I doing wrong when calculating manually.
Thanks for this comment, I was a bit concerned when he said that.
Thanks for the course Patrick! It was a great refresher!
BTW, at 3:42:02, in the newer versions instead of pretrained=True it is changed to weights=True.
This is one of the very few videos which is teaching Pytorch from the ground up! Beautiful work, @Python Engineer. Highly recommend it for any newbie + refresher.
Update: Note a subtle detail, if in with torch.no_grad() you use w = instead of w -= a new w variable will be created with requires_grad = False, which is fixed by w.requires_grad = True
Original: Using pytorch 1.11, and go figure @1:11 w.grad.zero_() errors, instead I had to put w.requires_grad = True
This is literally incredible. Perfect mix of theory and actual implementation. I can't thank you enough
.👆Girls dream of chatting with you
Wow this is so cool Patrick, a free course on PyTorch, great value you are bringing to the community 😆
Thanks so much :)
For the feedforward part, you need to send the model to the GPU when instantiating it:
model = NeuralNet(input_size, hidden_size, num_classes).to(device)
if your device is 'cuda' and you forget the '.to(device)' you will get an error.
omg thank you so much for this. saved me hours trying to figure out what was wrong serious life savor
at 1:01:41 he uses np.dot and when it should be np.multiply, that will make it consistent with the pytorch implementation. By doing np.dot, the items are multiplied and summed leaving just one value to which the mean function is applied, so the reason the numpy version get to 0 loss quicker is the gradient is not being averaged correctly.
thanks for pointing this out!
The best Pytorch tutorial online, I love how you explained the concepts using simple example and built on each concept one step at a time
I don't even need to watch it to know its quality. Can't wait to watch it and thanks for uploading!
Thanks! Hope you like it
I just completed the course on ML from scratch from Python Engineer. It was a great course for someone who learned all those algorithms in the past and wants to see how they get implemented using basic python lib and numpy.
On 4:14:00, I think you should use the ground truth as the labels rather than the predicted (line 130). Because the PR curve use the ground truth and predicted score to paint
Finally PyTorch doesnt seem as scary as it was before. The best tutorial I could find out there and I understood everything you've said. Thanks a lot.
glad to hear that :)
This is probably one of the best tutorials I've ever seen for pytorch. Thank you so much.
Thanks a lot! Glad you enjoy the course
If you guys get an error on GPU at around 3:13:50, saying there is two devices, make sure you do model.to(device)
Thanks a lot for the low level explanations.
At 1:01:47 when you dot product the array turns into a single scalar. So mean() returns that number(the sum), not average.
When you fix it you get the exact same results as with pytorch's implementation in 1:12:00
What is the correct expression of the gradient that gives the same result?
@@phi6934 I don't remember the details right now, but just dividing the expression with the size of the tensor must do the work. In the expression put smt like .../len(x) instead of .mean()
@@emrek1 yup that works thanks
I found that problem too, Thanks bro!
Best pytorch video tutorial I have found on entire internet. Also the codes are published. Just awesome
thanks a lot :)
Thanks for the awesome course! The material is extremely well curated, every minute is pure gold. I particularly liked the fact that for each subject there is a smooth transition from numpy to torch. It's perfect for someone who wants a quick and thorough deeplearning recap and get comfortable with hands-on pytorch coding.
The best hands-on tutorial on PyTorch on UA-cam! Thank you!
The man the myth the LEGEND returns with the best video of all time. 💪🏻
GREAT JOB and THANK YOU! ❤️
Thank you :)
Dear with apologies kindly notice, At timestamp 1:12:05 make a correction in stating, that the backprop grad was not correct, Actually the numerical one was not correct. Because np.dot is computing a single number and then taking mean is the same number, instead use 2*x/4 in np.dot(2*x,(Y_pred-Y).mean()) to correct your numerical gradient. Using np.dot(2*x/4,(Y_pred-Y)) will produce same result as back propagated result. Mean will be usefull when W and X are matrices.
Thank you
by FAR the best, most complete and comprehensible tutorial for pytorch I've come across
This is the best course on this topic I've seen so far. It is perfect when you want to understand what you're doing and the way things are brought is very pedagogic.
If z is a scalar then z.backward() is defined (and I understand the computation), while if z is not a scalar then z.backward() is not defined unless you provide appropriate inputs. However, it was not entirely clear to me what computation is occurring when we do z.backward(x) for example (where x is appropriate). This subject matter is around 33:00.
Same happened with me
What is happening is that PyTorch is assuming that you have provided the intermediate gradients i.e. (dLoss/dz), then using these intermediate gradients PyTorch is able to compute the gradients further downstream and backward step is successful.
amazing tutorial man! thank you so much !!! this is just the best!
This UA-cam video is the best tutorial for pytorch out there.Thankyou so much!
Wow, thanks!
OMG, you are an amazing teacher! Finally, I can grasp PyTorch and start building stuff. thank you so much
This vid quality is ridiculously high, THANK YOU
2:59:00 -> Starting with PyTorch 1.13 examples.next() is no longer valid.
New syntax is: next(examples)
The best PyTorch tutorials I've ever watched.
Best tutorial on pytorch I've come across.
Patrick, you're a legend. Thank you so much for this tutorial. Now on to more advanced stuff!
thanks a lot!
In the Gradient Descent and Training Pipeline sections, the presenter glosses over why it takes 5x more training steps to converge. There are a couple factors:
- Autograd is less aggressive than the manual gradient calculation, effectively lowering the learning rate (you can go all the way up to 0.1 after you move to torch and autograd)
- nn.Linear() includes a bias by default and a non-zero initialization of the weights, making it not a direct comparison. You can get much closer by adding `bias=False` to the model initialization and by zeroing out the weigth with `model.weight.data.fill_(0.0`
Someone has probably mentioned this already, but on line 23 at 1:04:08 .mean() is not doing anything since taking the dot product already returned a scalar. This is just dividing by one. Instead, you should be dividing by len(x) or len(y), or there may be another more efficient way to get the same result.
One of the best PyTorch tutorial series on UA-cam :)
this video was super helpful and clear, I watched everything up until transfer learning, ty so much
Thank you Python Engineer! This is the best tutorial video I've ever seen about pytorch.
1:12:09 It's because the gradient in your formula is not correct, not because pytorch's backpropogation calculation. You should put the ".mean()" into the brackets of "np.dot()".
Basic operations we can do, so x and y equals torch. so let's print x and y. So we do simple addition for example
41:01 Please change torch.optim.SGD(weights,lr=0.01) to torch.optim.SGD([weights],lr=0.01), here wights are passed as array
the most useful video I have ever watched
happy to hear that!
When you explained backprop, I felt like I finally saw the light at an endless tunnel
hehe, happy to hear that!
thank u for your patience!
I followed all courses and this helps me a lot. Thanks a ton
Thanks a lot, this tutorial helped me tremendously with my bachelors thesis
unbelievably excellent free tutorial course! Thank you!
Glad it was helpful!
This is the best Pytorch tutorial ever, thanks you!
Ten-soooor and Inter-ference are the best of the class!
Best course on pyTorch tutorial, thanks!
Your course is great! Congratulations!
I just had to do a small correction in your code in part "13. Feed Forward Net" so that I could run it on GPU. It was necessary to add the "device" (that was preciously declared) as an argument in the nn.Linear function. Without this detail it is not possible to run the code in GPU.
class NeuralNet(nn.Module):
def __init__(self, input_size, hidden_size, n_classes, device):
super(NeuralNet,self).__init__()
self.l1 = nn.Linear(input_size, hidden_size, device=device)
self.relu = nn.ReLU()
self.l2 = nn.Linear(hidden_size, n_classes, device=device)
def forward(self, x):
out = self.l1(x)
out = self.relu(out)
out = self.l2(out)
return out
Merci Beaucoup
Here's the best channel for data science and ML
Came for pytorch, stayed for the accent!
TENZSOoooOR 😎
haha :D
How is it that for a feed forward neural network we zero the gradients first before computing gradients and updating weights @3:08:35, whereas in the case of linear/logistic regression, we zero the gradients after computing them and updating the weights @1:36:19 @1:52:41.
Intuitively, this should not make any difference, but i wanted to confirm if that truely is the case. Is this just a nomenclature thingy?
Such a clear and comprehensive tut for Pytorch!
glad you like it :)
Really nice, well explained, well tested, etc.. Thanks a lot!!
Amazing and Comprehensive coverage of PyTorch. Amazing Video. Thanks a lot
This is the best tutorial on PyTorch
Nice tutorial !
@1:11:40 at line # 37. Instead of using "w -= learning_rate * w.grad" , I used expanded form "w = w - learning_rate * w.grad" and thought it would be same. But in this case 'w.grad' return 'None'. w.require_grad is False and hence error.
Though "w -= learning_rate * w.grad" is same as "w.data = w.data - learning_rate * w.grad".
It seems torch Tensor ( with require_grad True) have some overridden "__iadd__" implementation.
unsupported operand type(s) for *: 'float' and 'builtin_function_or_method' got this error on that line. any help please
Man this is pure gold, thank you so much!
I finished the whole video, again, thank you so much!
Note at 2:08: `dataiter.next()` is no throwing an AttributeError: '_MultiProcessingDataLoaderIter' object has no attribute 'next'... I changed that line to `data = next(dataiter)`
I have just finished the whole tutorial as a refresher. Everything is so much clearer now. Thanks.
Thankyou Patrick. It was a fantastic tutorial.
Thank you very much! literally the best place to learn pytorch
basic explanation about autograd was great
Absoulte top quality videos! Thank you very much and may you go on forever
This is an error I have found
Time: 1:01:55
According to the equation,we actually need to find 1/N ,where N represents the number of term(here 4).According to the code,we are computing mean after converting the rest of the code to a dot product,which contains just a value.So instead of dividing with the desired value(4),we are dividing with 1.
This is amazing! It was fun to follow along and I feel like I am able to try pytorch on some projects now. Thank you 😍
At 1:01:55 you are taking the mean of a scalar, which doesn't do anything. Since you have 4 data points only this effectively means that your learning_rate was multiplied by 4. This is the reason why it seems to work better than PyTorch: this particular case is so well behaved that to speed up is sufficient to take larger steps.
This tutorial is supppppppppppper great! The best deep learning tutorial I've ever watched. Thank you so much.
I enjoined the tutorial that I didn't want it to stop!
I look forward to seeing more great videos like this from this channel
Awesome, thank you!
Thanks for this incredible resource. FYI I believe the gradient function computed at 1:01:38 is incorrect. I'm pretty sure it should be:
def gradient(x, y, y_predicted):
return ((y_predicted-y)*2*x).mean()
Finished the tutorial love it
a probable mistake: Leaky ReLU isn't used for solving the problem of vanishing gradient problem but Dead Neurons problem. Which can happen when you use ReLU activation functions.
Dec. 1st 7:38
Dec. 2nd 1:02:30
Perfect tutorial for a beginner!!!!!!!!
Glad you think so!
best pytorch tutorial ever
Thank you so much, if I got a job by watching this, I want to make a donation.
2:58:47 examples.next() doesn't work for me. Instead use next(examples)
Thank you so much, I had that problem at 2:07:54
such a brilliant course !! I thank you so much !!
This course is amazing !! Thanks of everythink.
At 04:40 I needed to open the Anacoda terminal because it didnt recognise the 'conda' comand on the windows terminal.
Ah yes, on windows you have to either manually add it to your PATH, or simply use the Anaconda terminal
Thankyou Patric for your Fantastic tutorial. ☺
Dude this has general helped me so much. Thank you!
Glad to hear it!
thanks for the detailed tutorial. Subscribed.
Thank you for the tutorial, i enjoyed it
Absolutely great. But what was missing for me was how then to use a trained model. Conspicuous in its absence was how at the end to feed data into a trained model and get the answer it was trained to give. Is there another video that explains this?
Wanna explore a package like pytorch? run print(dir(torch)) or any other package/module and you'll get an interesting printout of available functions.
Great course, as always!
Well done, a very smooth intro to PyTorch.
Glad you like it!
1:42:10 Is there a reason you used `X, y` instead of `X, Y`? I believe it should `Y` as we're dealing with a tensor of dependent variables right? It would be `y` if we were dealing with a scalar though
thank you for the great video. I learnt so much from you!
Cool, really a very nice course, thanks for your effort to make it free online!!!
Glad you enjoyed it!
Thanks for your video, it's so nice!
I've taken a graduate course in deep learning and neural, and have watched other tutorials here and there, but this is by far the most helpful one. Granted, all the previous materials have probably contributed, but the way you teach is unparalleled!
thank you so much! glad you like it :)
That is an excellent course. Thank you Python Engineer
Very good tutorial, good job, thank you for this course!!
wow, it's really help a lot! Thanks for sharing!
great course. thanks so much for sharing your knowledge