Torch.load() is behaving strangely

require 'torch'   -- torch
require 'optim'
require 'nn'      -- provides a normalization operator

function string:split(sep)
  local sep, fields = sep, {}
  local pattern = string.format("([^%s]+)", sep)
  self:gsub(pattern, function(substr) fields[#fields + 1] = substr end)
  return fields

local f_file =[1], 'r')
local p_file =[2], 'w')
local data = torch.Tensor(1, 351)
local name = ''
for line in f_file:lines('*l') do
local l = line:split(',')
first = true
	for key, val in ipairs(l) do
	  if first == false then
	  data[1][key] = val
	  else data[1][key] = 0
	  first = false
	  name = val

local X = data[{{},{2,-1}}]

 model = torch.load('estimation_model.dat')

 myPrediction = model:forward(X)


```my model is not being loaded,I think its being in .dat format might be the problem .
Here is the erro
th> model=torch.load(tracking_model.dat)
[string "model=torch.load(tracking_model.dat)"]:1: attempt to index global 'tracking_model' (a nil value)
stack traceback:
	/home/sunny/torch/install/share/lua/5.2/trepl/init.lua:506: in function '__index'
	[string "model=torch.load(tracking_model.dat)"]:1: in main chunk
	[C]: in function 'xpcall'
	/home/sunny/torch/install/share/lua/5.2/trepl/init.lua:679: in function 'repl'
	...unny/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:204: in main chunk
	[C]: in ?

It looks like you are using LuaTorch, which is not in active developement anymore.
Are you using an older code base and would it be possible to port it to PyTorch?