Xin (Simon) Dong XinDongol

GeForce GTX 1080 Ti Mem: 12GB

processor : 15 vendor_id : GenuineIntel cpu family : 6 model : 63 model name : Intel(R) Core(TM) i7-5960X CPU @ 3.00GHz stepping : 2 microcode : 0x3d

	from graphviz import Digraph
	from torch.autograd import Variable
	import torch


	def make_dot(var, params=None):
	if params is not None:
	assert isinstance(params.values()[0], Variable)
	param_map = {id(v): k for k, v in params.items()}

	import torch.nn as nn
	import torch.nn.functional as F
	import torch
	class DeepInversionFeatureHook():
	'''
	Implementation of the forward hook to track feature statistics and compute a loss on them.
	Will compute mean and variance, and will use l2 as a loss
	'''
	def __init__(self, module):
	self.hook = module.register_forward_hook(self.hook_fn)

	\usepackage{soul}
	\newcommand{\xinc}[1]{\textcolor{blue}{[Xin:$\rightarrow$#1]}}
	\newcommand{\xina}[1]{\textcolor{blue}{[Xin:$+$#1]}}
	\newcommand{\xin}[1]{\textcolor{blue}{[Xin:#1]}}
	\newcommand{\xind}[1]{\textcolor{blue}{[Xin:\st{#1}]}}

	%% parameters
	delta = 0.5;
	a2 = 0.1;
	a1 = 100;
	number_of_pieces = 100; % number of pieces we want to divide
	%%
	b2 = delta/tanh(a2*delta);
	b1 = delta/tanh(a1*delta);
	thres_min = zeros(1,number_of_pieces)+1000;
	amin = zeros(1,number_of_pieces);

	from torch.autograd import Variable
	import torch
	from torch import nn
	from collections import OrderedDict
	from IPython import embed
	from torch.autograd.function import InplaceFunction, Function
	import torch.nn.functional as F
	import math