torch学习笔记一（基本语法）

最新推荐文章于 2024-08-04 15:12:08 发布

why小曼

最新推荐文章于 2024-08-04 15:12:08 发布

阅读量4.4k

点赞数 1

分类专栏： torch学习笔记文章标签： torch学习

本文链接：https://blog.csdn.net/Sophie94/article/details/52681195

版权

torch学习笔记专栏收录该内容

3 篇文章 0 订阅

订阅专栏

一 Lua基本语法：

1 变量：

 1. 变量默认都是全局变量（一个文件中的变量所有的文件都可以访问），加上local变为局部变量，仅仅该文件可访问。
 2. 变量类型：nil、Boolean、string、Number和table 。 在Lua中使用变量不需要提前声明，变量的类型决定于用户赋值的类型 。 可以使用 type()函数判断变量的类型。
 3. 给一个变量赋值为nil，表示释放该变量。Boolean跟其他语言一样，只有true和false两种值。Number是双精度浮点数，Lua中没有整数类型。
 4. String可以用单引号或者双引号来初始化，连接使用两个字符串使用..
 5.  未定义的 变量值为nil，不是错误。只有nil 和false为假，0 和 '' 为真

2 注释：
用 “- - ”来标记该行的注释，使用“- - [ [” 和 “ - - ] ] ”之间括起来的部分进行块注释
3 运算符：
Lua中支持的算术运算符有：+、-、*、/,支持的关系运算符有：==、~=(不等于)、<、>、<=、>=；支持的逻辑运算符有：and、or、not，并且都是左短路
4 控制语句：

 1. if，else 中间多了then
 2. 代码块必须要有end结尾
 3. for循环中:begin,end,step的结构，包括begin和end

-- Blocks are denoted with keywords like do/end:
while num < 50 do
  num = num + 1  -- No ++ or += type operators.
end

-- If clauses:
if s ~= 'walternate' then  -- ~= is not equals.
  io.write('not over 40\n')  -- Defaults to stdout.
else
  local line = io.read()  -- Reads next stdin line.
  print('Winter is coming, ' .. line)
end

karlSum = 0
for i = 1, 100 do  -- The range includes both ends.
  karlSum = karlSum + i
end

-- Use "100, 1, -1" as the range to count down:
fredSum = 0
for j = 100, 1, -1 do fredSum = fredSum + j end

-- In general, the range is begin, end[, step].
repeat
  print('the way of the future')
  num = num - 1
until num == 0

5 函数
函数是以function关键字开始，然后是函数名称，参数列表，最后以end关键字表示函数结束。需要注意的是，函数中的参数是局部变量，如果参数列表中存在(…)时，Lua内部将创建一个类型为table的局部变量arg，用来保存所有调用时传递的参数以及参数的个数(arg.n) 。

function fib(n)
  if n < 2 then return 1 end
  return fib(n - 2) + fib(n - 1)
end

-- Closures and anonymous functions are ok:
function adder(x)
  -- The returned function is created when adder is
  -- called, and remembers the value of x:
  return function (y) return x + y end
end
a1 = adder(9)
a2 = adder(36)
print(a1(16))  --> 25
print(a2(64))  --> 100

-- Unmatched receivers are nil;
-- unmatched senders are discarded.
function bar(a, b, c)
  print(a, b, c)
  return 4, 8, 15, 16, 23, 42
end

x, y = bar('zaphod')  
--> prints "zaphod  nil nil"
-- Now x = 4, y = 8, values 15..42 are discarded.

-- Functions are first-class, may be local/global.
-- These are the same:
function f(x) return x * x end
f = function (x) return x * x end

-- And so are these:
local function g(x) return math.sin(x) end
local g; g  = function (x) return math.sin(x) end
-- the 'local g' decl makes g-self-references ok.
-- Trig funcs work in radians, by the way.

6 表
table可以作为列表或者字典使用，使用非nil的值作为key,通常是字符或者数字

t = {key1 = 'value1', key2 = false}
print(t.key1)  -- Prints 'value1'.
t.newKey = {}  -- Adds a new key/value pair.
t.key2 = nil   -- Removes key2 from the table.

-- Key matching is basically by value for numbers and strings, but by identity for tables.
u = {['@!#'] = 'qbert', [{}] = 1729, [6.28] = 'tau'}
a = u['@!#']  -- Now a = 'qbert'.
b = u[{}]     -- We might expect 1729, but it's nil:
-- b = nil since the lookup fails because the key we used is not the same object as the one used to store the original value. 

for key, val in pairs(u) do  -- Table iteration.
  print(key, val)
end

-- _G is a special table of all globals.
print(_G['_G'] == _G)  -- Prints 'true'.

-- List literals implicitly set up int keys:
v = {'value1', 'value2', 1.21, 'gigawatts'}
for i = 1, #v do  -- #v is the size of v for lists.
  print(v[i])  
end
-- A 'list' is not a real type. v is just a table

7 类：
通过table实现类的功能

-- Classes aren't built in; there are different ways
-- to make them using tables and metatables.

Dog = {}                                   -- 1.

function Dog:new()                         -- 2.
  newObj = {sound = 'woof'}                -- 3.
  self.__index = self                      -- 4.
  return setmetatable(newObj, self)        -- 5.
end

function Dog:makeSound()                   -- 6.
  print('I say ' .. self.sound)
end

mrDog = Dog:new()                          -- 7.
mrDog:makeSound()  -- 'I say woof'         -- 8.

-- 1. Dog acts like a class; it's really a table.
-- 2. function tablename:fn(...) is the same as
--    function tablename.fn(self, ...)
--    The : just adds a first arg called self.
--    Read 7 & 8 below for how self gets its value.
-- 3. newObj will be an instance of class Dog.
-- 4. self = the class being instantiated. Often
--    self = Dog, but inheritance can change it.
--    newObj gets self's functions when we set both
--    newObj's metatable and self's __index to self.
-- 5. Reminder: setmetatable returns its first arg.
-- 6. The : works as in 2, but this time we expect
--    self to be an instance instead of a class.
-- 7. Same as Dog.new(Dog), so self = Dog in new().
-- 8. Same as mrDog.makeSound(mrDog); self = mrDog.

-- Inheritance example:

LoudDog = Dog:new()                           -- 1.

function LoudDog:makeSound()
  s = self.sound .. ' '                       -- 2.
  print(s .. s .. s)
end

seymour = LoudDog:new()                       -- 3.
seymour:makeSound()  -- 'woof woof woof'      -- 4.

-- 1. LoudDog gets Dog's methods and variables.
-- 2. self has a 'sound' key from new(), see 3.
-- 3. Same as LoudDog.new(LoudDog), and converted to
--    Dog.new(LoudDog) as LoudDog has no 'new' key,
--    but does have __index = Dog on its metatable.
--    Result: seymour's metatable is LoudDog, and
--    LoudDog.__index = LoudDog. So seymour.key will
--    = seymour.key, LoudDog.key, Dog.key, whichever
--    table is the first with the given key.
-- 4. The 'makeSound' key is found in LoudDog; this
--    is the same as LoudDog.makeSound(seymour).

-- If needed, a subclass's new() is like the base's:
function LoudDog:new()
  newObj = {}
  -- set up newObj
  self.__index = self
  return setmetatable(newObj, self)
end

8 模块

require文件会被缓存，所以一般最多只调用一次require;dofile 类似require方式，但是没有缓存，可以多次调用;loadfile只load文件而没有运行文件
‘：’和’.’的区别：’只要出现了”:”，对于调用就将自身作为第一个参数传入；对于函数定义的话，就将传入的第一个参数作为self。obj.Wakeup(obj,1)和obj:Wakeup(1)是等价的。结果都是传入2个参数：obj和1 。

-- Suppose the file mod.lua looks like this:
local M = {}
local function sayMyName()
  print('Hrunkner')
end
function M.sayHello()
  print('Why hello there')
  sayMyName()
end
return M

-- Another file can use mod.lua's functionality:
local mod = require('mod')  -- Run the file mod.lua.

-- require is the standard way to include modules.
-- require acts like:     (if not cached; see below)
local mod = (function ()
  <contents of mod.lua>
end)()
-- It's like mod.lua is a function body, so that
-- locals inside mod.lua are invisible outside it.

-- This works because mod here = M in mod.lua:
mod.sayHello()  -- Says hello to Hrunkner.

-- This is wrong; sayMyName only exists in mod.lua:
mod.sayMyName()  -- error

-- require's return values are cached so a file is
-- run at most once, even when require'd many times.

-- Suppose mod2.lua contains "print('Hi!')".
local a = require('mod2')  -- Prints Hi!
local b = require('mod2')  -- Doesn't print; a=b.

-- dofile is like require without caching:
dofile('mod2.lua')  --> Hi!
dofile('mod2.lua')  --> Hi! (runs it again)

-- loadfile loads a lua file but doesn't run it yet.
f = loadfile('mod2.lua')  -- Call f() to run it.

-- loadstring is loadfile for strings.
g = loadstring('print(343)')  -- Returns a function.
g()  -- Prints out 343; nothing printed before now.

参考资料：
http://tylerneylon.com/a/learn-lua/

二 Tensor：

1 表示 Tensor/rand/zeros/ones

a = torch.Tensor(2,3)
b = torch.zeros(2,2)
t = torch.ones(2,2)
c = torch.rand(2,2)

2 转换：floor/ceil/int/long/zero/fill

a = torch.rand(2,2)
b = a:ceil():int():zero()

3 查询尺寸：size/nDimension/nElement

s = torch.LongStorage(6)
s[1] = 4; s[2] = 5; s[3] = 6; s[4] = 2; s[5] = 7; s[6] = 3;
x = torch.Tensor(s)
> x:nDimension()
6
> x:size()
 4
 5
 6
 2
 7
 3
[torch.LongStorage of size 6]
> x:nElement()
5040

4 运算：equal/add/mul/cmul/
四种表达形式：
y = torch.add(a, b) returns a new Tensor.
torch.add(y, a, b) puts a + b in y.
a:add(b) accumulates all elements of b into a.
y:add(a, b) puts a + b in y。
add 操作时，元素的数目必须match，但是size可以不match
mul是tensor与数字的乘法，而cmul则是tensor之间的乘法

--Returns true if the dimensions and values of tensor1 and tensor2 are exactly the same.
a = torch.zeros(1,2)                                                                    
b = torch.ones(1,2)
> a:equal(b)
false
> x = torch.Tensor(2, 2):fill(2)
> y = torch.Tensor(4):fill(3)
> x:add(y)
> x
 5  5
 5  5
[torch.DoubleTensor of size 2x2]
x:add(2, y) --multiply y with 2
> x
 11  11
 11  11
[torch.DoubleTensor of size 2x2]
> x:mul(2)
 22 22
 22 22
>x:cmul(y) 
 66 66
 66 66

5 截取：select/narrow/sub/{}

select是直接提取某一维；narrow是取出某一维来截取一部分； sub就是取出一块，是对取出的所有维进行裁剪。
语法格式：语法： select(dim, index)； narrow(dim, index, num); sub(dim1s, dim1e, dim2s, dim2e,…)
用”{ }和[]”来提取
[ {dim1 , dim2, …} ]来获取某些维度。类似select
[ { {dim1s, dim1e}, {dim2s, dim2e},… } ] 类似sub
[ { dim1, {dim2s, dim2e},… }]相当于narrow

x = torch.Tensor(3,4)
i = 0 
x:apply(function()i = i+1 return i end)
--[[
x 为
  1   2   3   4
  5   6   7   8
  9  10  11  12
]]
selected = x:select(1,2)  --第一维的第二个。就是第二行。相当于x[2]
narrowed = x:narrow(2,1,2)
--[[
th> narrowed
  1   2
  5   6
  9  10
]]
subbed = x:sub(1,3,2,3)
--[[ 一维到3为止，二维也到3为止。
th> subbed
  2   3
  6   7
 10  11
]]
x[{1,2}] -- 2
x[{1,{1,2}}] --1 2