PyTorch fundamentals exercise

[NEELANATARAJAN]

Question 4: Set the random seed to 0 and do 2 & 3 over again.
The output should be:

(tensor([[1.8542],
[1.9611],
[2.2884],
[3.0481],
[1.7067],
[2.5290],
[1.7989]]), torch.Size([7, 1]))

If we initiate the seed using the manual_seed command and create two tensors and matmul with 2nd tensor transposed we are getting the desired result

But as mentioned in the course if the manual_seed command is called before creating the input tensors separately, the results vary

Can someone explain which is the right way to handle reproducibility? Should seed command be initiated before every tensor creation or only once before all the commands?

[NEELANATARAJAN]

Author has called twice separately in two different cells in the lecture. I did the same and the results do not match hence the query.

[37Lime]

Hello,

A good practice to obtain reproducibility is to set a seed at the beginning of your code (top of your notebook cell) and stick with the same seed number in your experiments.

Best wishes.