alkalait

import torch  # 1.3.1
import torch.nn as nn
import torchvision  # 0.4.2

# for example
resnet = torchvision.models.resnet50(pretrained=False)
# strip away its FC layer
resnet_noFC = nn.Sequential(*list(resnet.children())[:-1])

# Suppose you want to use it for binary classification.

alkalait

import torch
import numpy

def shape(name : str) -> None:
    '''
    yaarr = torch.ones(25, 1, 1, 5, 5)  # or np.ones
    shape('yaarr')
    >> yaarr : (25, 1, 1, 5, 5)
    '''
    val = eval(name)

alkalait

# Author: Freddie Kalaitzis
# License: MIT
# Source: https://gist.github.com/alkalait/1497032fb601997efd9be4b90dddc63b

import matplotlib.pyplot as plt
import numpy as np
import seaborn as sns
import torch
import xarray as xr
sns.set_style('white')

alkalait

from torch import Tensor

def is_broadcastable(x: Tensor, y: Tensor) -> bool:
    """ Are the shapes of two tensors compatible for broadcasting? """
    if not x.ndim == y.ndim:
        return False
    n_same_dim = (torch.as_tensor(x.shape) == torch.as_tensor(y.shape)).sum()
    return (int(n_same_dim) - x.ndim) <= 1

alkalait

# Author: Freddie Kalaitzis
# License: MIT
# Source: https://gist.github.com/alkalait/c99213c164df691b5e37cd96d5ab9ab2

import functools
import itertools
import os
import re
import warnings

alkalait

# batch, time, bands, height, width
B, T, C, H, W = x.shape
C_o = 32  # hyperparam

F = lambda i, o, g=1:
    nn.Conv2d(in_channels=i,
              out_channels=o,
              kernel_size=3, 
              padding='same',
              groups=g)