Bruce_Ho
September 8, 2018, 10:50pm
1
I am not sure this is a pytorch question or a python question. In
import numpy as np
import torch
from torch.autograd import Variable
import torch.optim as optim
import math
from tqdm import tqdm
from torch.utils.data.dataset import Dataset
from torch.utils.data import DataLoader
from .lda2vec_loss import loss, topic_embedding
# negative sampling power
BETA = 0.75
# i add some noise to the gradient
ETA = 0.4
# i believe this helps optimization.
# the idea is taken from here:
# https://arxiv.org/abs/1511.06807
# 'Adding Gradient Noise Improves Learning for Very Deep Networks'
show original
line 127, why is it ok to call model(doc_indices, pivot_words, target_words) without specifying forward() method name?
1 Like
SimonW
September 8, 2018, 11:02pm
2
python question.
When you call model(...), you are actually calling model.__call__(...). As you can see here , the __call__ method on nn.Module eventually calls forward along with taking care of tracing and hooks.
5 Likes
chsafouane
August 26, 2019, 7:23pm
3
Would you mind explaining what do you by ‘tracing and hooks’ ?