[FIXED] Converting from list of very small floats to a PyTorch tensor gives a TypeError because PyTorch interprets scientific notation as str instead of float

Issue

I have data that I want to setup as a PyTorch tensor as my objective for a ML algorithm. If I inspect the list as is I get :

ydata = [1.39e-11, 1.5e-12, 4.7e-16]

I want to have this list in a tensor, so I tried multiple functions but everytime I get a TypeError for instance :

torch.FloatTensor(ydata) -> TypeError('must be real number, not str')

However if I inspect the type of each element of ydata it is a float. How can I go about getting the data I want into a tensor?

Thank you very much.

Solution

I just tried to replicate your example on my machine:

pip install torch torchvision torchaudio

then:

import torch
ydata = ['1.39e-11', '1.5e-12', '4.7e-16']
ydata = [float(x) for x in ydata]   # convert your string to float
torch.FloatTensor(ydata)

#output
tensor([1.3900e-11, 1.5000e-12, 4.7000e-16])

Answered By - Goku - stands with Palestine