rohitdavas

rohitdavas

loaded from https://github.com/philc/vimium/blob/master/README.md

?       show the help dialog for a list of all available keys
h       scroll left
j       scroll down
k       scroll up
l       scroll right
gg      scroll to top of the page
G       scroll to bottom of the page
d       scroll down half a page

rohitdavas

# the package you need is pipreqs
pip install pipreqs 

# they have a number of arguments. the most useful i found
# is the argument allowing you to check your package on pypi server. 
# this fixed bugs in the requriements.txt packages. e.g skimage is not a package, but pipreqs was listing it down
# another is the version number listed by the pipreqs might be messed up. 
python -m  pipreqs.pipreqs . --force --mode gt --pypi-server https://pypi.org/

rohitdavas

"""
This is simple implementation of a binary cross entropy.

Test cases : done.

"""
import torch 

def balanced_BCE_loss(predictions: torch.Tensor, ground_truth: torch.Tensor, with_logits=False) -> torch.Tensor:
    """

rohitdavas

# %%
import torch 
import torch.nn.functional as F # for using softmax
cc = torch.manual_seed(0) # for reproducibility
# somehow on my m1 mac, the randomness is not reproducible. todo: figure out why

b = 1 # mini batch size
t = 3 # sequence length
k = 2 # dimension of each vector in the sequence