Video "PyTorch for Deep Learning & Machine Learning – Full Course"- Coding error in tensor-like

[Rohi2610]

Hello Sir, I'm a student trying to learn PyTorch from your amazing YouTube video. As you suggested, I've been experimenting around with almost every topic that's being discussed. So after learning Tensor-like, I tried creating a Torch.rand_like. What happened here is that if I use torch.range() to create a tensor and then create a tensor-like, the code works smoothly.

Code:

one_ten= torch.range(1, 10)
ten_rand= torch.rand_like(input=one_ten) 
ten_rand

The error happens when I use torch.arange() or any tensor created using arange. I have attached the error as a picture. It basically says "check_uniform_bounds" not implemented for 'Long'. I tried changing the datatype of the tensor to float as some articles in Google and the collab AI suggested but it was no good. Below is the code with torch.arange().

one_ten= torch.arange(1, 10)
one_ten.float
ten_rand= torch.rand_like(input=one_ten) 
ten_rand

[LuluW8071]

You need to change torch.dtype() when using torch.arange() since in the deprecated (old version) of Torch, when using torch.range(), the values were float32 by default. Now, for torch.arange(), its dtype is Long by default.

So, looking at your RuntimeError: "check_uniform_bounds" not implemented for 'Long' error, one_ten generated by torch.arange() yields Long values, but torch.rand_like only accepts float values. To debug this, simply change the dtype.

one_ten= torch.arange(1, 10).to(torch.float32)
ten_rand= torch.rand_like(input=one_ten) 
ten_rand

You can also do print(one_ten) on both range and arange to know the difference and check their dtypes.

[Rohi2610]

Thank you so much. It worked and I understood the problem.