How to Print and Convert Google Takeout to PDF?

The following post will provide an excellent and efficient method to Convert Google Takeout to PDF. It entails the use of a Google Takeout Emails application as a PDF converter. The majority of people are looking for a Google Takeout to PDF Converter to help them evaluate their Gmail correspondence. Google Takeout is a free service that allows users to back up all of their data from Google services like Gmail, YouTube, G Drive, Calendars, and more.

All of the data is saved in several file formats, with Gmail emails being saved in the MBOX format. MBOX stands for mailbox, which is also a well-known file format nowadays. Unfortunately, not all email clients or services support MBOX. You'll do this by converting it to a PDF file, which is both flexible and cross-platform compatible.

The most basic reason why people desire to convert Google Takeout emails to PDF is that PDF files are the most secure files available, since they are independent of operating systems and can be accessed on any device. It's difficult to manually convert data to the flexible file format because there isn't a simple way to do it.

You will learn how to Import Google Takeout to PDF converter and how it may assist users in the required field. But first, let us offer you a clear picture of why a PDF file is an excellent option for saving emails.

While Google Takeout allows you to download and save Google data to your computer, it does not support your preferred file types. If you wish to export Takeout emails, for example, MBOX is the only format you have (also contained within ZIP file). As a result, e-mails must be converted to the required format using Google Takeout's reliable conversion program.

The fastest and most dependable Google Takeout to PDF Converter method is provided below, which exports all emails in the Google Takeout architecture to PDF format quickly. It also saves all attachments and offers a variety of other customization. If the software meets your needs, you may want to test it out using the free demo version.

Why do users want to PDF Google Takeout?

There are several reasons why a user wants to convert their data. In the next section you will find some of the advantages of having a PDF file. Check out these.

1. PDF file formats are not platform-dependent and are available on all main devices.
2. It maintains data integrity with other qualities regardless of the OS on which it is read.
3. Another wonderful reason to search out for a Google take-out PDF converter is the fact that it is the most secure and safe file format with password security and numerous other levels of permissions.
4. This is the only lawful file format that works in courts.
5. PDF is a standard file format that is accepted globally.

The Ideal Solution for Converting Google Takeout to PDF.

If you're wondering how to save Gmail archive emails as PDF files, then you're in the right place. This blog is for those users who are looking for a way to save their downloaded Gmail emails into PDF files.

In today's digital age, consumers frequently migrate their email data from one file type to another for security reasons. That is why we are providing an expert-recommended solution, namely SoftwareImperial Converter for Google Takeout. It is a program that allows you to save Google Takeout files in the PDF file format, complete with attachments. Because when you use Google's takeout service to download emails from your Gmail account, Google saves all of your data in a ZIP folder. Additionally, when you unzip the ZIP file, you will receive a .mbox file containing all of your Gmail emails. As a result, the software is an advantageous alternative for swiftly saving Gmail archive emails as PDF files. Now, let's have a look at the program's advantageous features and also discover how to save a Google Takeout file to PDF using this application.

Features of the Save Gmail Archive Software Email as PDF

It offers the simultaneous storage of single Gmail and bulk emails in PDF format.
It allows you to see every email from Google Takeout before converting to PDF format.
It is suitable to export all MBOX file versions – with extensions and without extensions such as .mbox, MBOX, .MBS, and .MBX.
This offers powerful filters to preserve chosen PDF emails from Gmail archives.
Save all emails in a PDF file. 2. Save individual PDF files for each email and 3. Save PDF files as a non-searchable PDF. The software provides three alternative PDF saving choices, 1.
It also contains many attachment saving choices such as: 1. Add attachments to the PDF directly, 2. Save attachments into folders separately, 3. Turn all attachments into PDF formats, and 4. Add attachments in a PDF document. You can choose any choice according to your needs.
It is an application for saving Gmail archive mails in PDF format based on Windows. This utility can be run on Windows 10, 8.1, 8, 7, XP, etc.

Synopsis:

The PDF files are often chosen, as they are available and accessible on all email and operating systems platforms. Because it is an independent and versatile file format, users can also find methods to Import Google Takeout to PDF in an absolutely safe file format.

Although there are no direct manual solutions, we have provided you with the only direct converting option, which is professional software. Make the most of the Google Takeout to PDF Converter and understand everything it offers for a wonderful conversion experience.

Relevant Informative Articles

How to Export Gmail to Outlook?

Solution about How to Import Google Takeout to New Account?

What is GEEK

Buddha Community

How to Print and Convert Google Takeout to PDF?

Tamale Moses

1669003576

Exploring Mutable and Immutable in Python

In this Python article, let's learn about Mutable and Immutable in Python.

Mutable and Immutable in Python

Mutable is a fancy way of saying that the internal state of the object is changed/mutated. So, the simplest definition is: An object whose internal state can be changed is mutable. On the other hand, immutable doesn’t allow any change in the object once it has been created.

Both of these states are integral to Python data structure. If you want to become more knowledgeable in the entire Python Data Structure, take this free course which covers multiple data structures in Python including tuple data structure which is immutable. You will also receive a certificate on completion which is sure to add value to your portfolio.

Mutable Definition

Mutable is when something is changeable or has the ability to change. In Python, ‘mutable’ is the ability of objects to change their values. These are often the objects that store a collection of data.

Immutable Definition

Immutable is the when no change is possible over time. In Python, if the value of an object cannot be changed over time, then it is known as immutable. Once created, the value of these objects is permanent.

List of Mutable and Immutable objects

Objects of built-in type that are mutable are:

Lists
Sets
Dictionaries
User-Defined Classes (It purely depends upon the user to define the characteristics)

Objects of built-in type that are immutable are:

Numbers (Integer, Rational, Float, Decimal, Complex & Booleans)
Strings
Tuples
Frozen Sets
User-Defined Classes (It purely depends upon the user to define the characteristics)

Object mutability is one of the characteristics that makes Python a dynamically typed language. Though Mutable and Immutable in Python is a very basic concept, it can at times be a little confusing due to the intransitive nature of immutability.

Objects in Python

In Python, everything is treated as an object. Every object has these three attributes:

Identity – This refers to the address that the object refers to in the computer’s memory.
Type – This refers to the kind of object that is created. For example- integer, list, string etc.
Value – This refers to the value stored by the object. For example – List=[1,2,3] would hold the numbers 1,2 and 3

While ID and Type cannot be changed once it’s created, values can be changed for Mutable objects.

Check out this free python certificate course to get started with Python.

Mutable Objects in Python

I believe, rather than diving deep into the theory aspects of mutable and immutable in Python, a simple code would be the best way to depict what it means in Python. Hence, let us discuss the below code step-by-step:

#Creating a list which contains name of Indian cities

cities = [‘Delhi’, ‘Mumbai’, ‘Kolkata’]

# Printing the elements from the list cities, separated by a comma & space

for city in cities:
		print(city, end=’, ’)

Output [1]: Delhi, Mumbai, Kolkata

#Printing the location of the object created in the memory address in hexadecimal format

print(hex(id(cities)))

Output [2]: 0x1691d7de8c8

#Adding a new city to the list cities

cities.append(‘Chennai’)

#Printing the elements from the list cities, separated by a comma & space

for city in cities:
	print(city, end=’, ’)

Output [3]: Delhi, Mumbai, Kolkata, Chennai

#Printing the location of the object created in the memory address in hexadecimal format

print(hex(id(cities)))

Output [4]: 0x1691d7de8c8

The above example shows us that we were able to change the internal state of the object ‘cities’ by adding one more city ‘Chennai’ to it, yet, the memory address of the object did not change. This confirms that we did not create a new object, rather, the same object was changed or mutated. Hence, we can say that the object which is a type of list with reference variable name ‘cities’ is a MUTABLE OBJECT.

Let us now discuss the term IMMUTABLE. Considering that we understood what mutable stands for, it is obvious that the definition of immutable will have ‘NOT’ included in it. Here is the simplest definition of immutable– An object whose internal state can NOT be changed is IMMUTABLE.

Again, if you try and concentrate on different error messages, you have encountered, thrown by the respective IDE; you use you would be able to identify the immutable objects in Python. For instance, consider the below code & associated error message with it, while trying to change the value of a Tuple at index 0.

#Creating a Tuple with variable name ‘foo’

foo = (1, 2)

#Changing the index[0] value from 1 to 3

foo[0] = 3
	
TypeError: 'tuple' object does not support item assignment

Immutable Objects in Python

Once again, a simple code would be the best way to depict what immutable stands for. Hence, let us discuss the below code step-by-step:

#Creating a Tuple which contains English name of weekdays

weekdays = ‘Sunday’, ‘Monday’, ‘Tuesday’, ‘Wednesday’, ‘Thursday’, ‘Friday’, ‘Saturday’

# Printing the elements of tuple weekdays

print(weekdays)

Output [1]:  (‘Sunday’, ‘Monday’, ‘Tuesday’, ‘Wednesday’, ‘Thursday’, ‘Friday’, ‘Saturday’)

#Printing the location of the object created in the memory address in hexadecimal format

print(hex(id(weekdays)))

Output [2]: 0x1691cc35090

#tuples are immutable, so you cannot add new elements, hence, using merge of tuples with the # + operator to add a new imaginary day in the tuple ‘weekdays’

weekdays  +=  ‘Pythonday’,

#Printing the elements of tuple weekdays

print(weekdays)

Output [3]: (‘Sunday’, ‘Monday’, ‘Tuesday’, ‘Wednesday’, ‘Thursday’, ‘Friday’, ‘Saturday’, ‘Pythonday’)

#Printing the location of the object created in the memory address in hexadecimal format

print(hex(id(weekdays)))

Output [4]: 0x1691cc8ad68

This above example shows that we were able to use the same variable name that is referencing an object which is a type of tuple with seven elements in it. However, the ID or the memory location of the old & new tuple is not the same. We were not able to change the internal state of the object ‘weekdays’. The Python program manager created a new object in the memory address and the variable name ‘weekdays’ started referencing the new object with eight elements in it. Hence, we can say that the object which is a type of tuple with reference variable name ‘weekdays’ is an IMMUTABLE OBJECT.

Also Read: Understanding the Exploratory Data Analysis (EDA) in Python

Where can you use mutable and immutable objects:

Mutable objects can be used where you want to allow for any updates. For example, you have a list of employee names in your organizations, and that needs to be updated every time a new member is hired. You can create a mutable list, and it can be updated easily.

Immutability offers a lot of useful applications to different sensitive tasks we do in a network centred environment where we allow for parallel processing. By creating immutable objects, you seal the values and ensure that no threads can invoke overwrite/update to your data. This is also useful in situations where you would like to write a piece of code that cannot be modified. For example, a debug code that attempts to find the value of an immutable object.

Watch outs: Non transitive nature of Immutability:

OK! Now we do understand what mutable & immutable objects in Python are. Let’s go ahead and discuss the combination of these two and explore the possibilities. Let’s discuss, as to how will it behave if you have an immutable object which contains the mutable object(s)? Or vice versa? Let us again use a code to understand this behaviour–

#creating a tuple (immutable object) which contains 2 lists(mutable) as it’s elements

#The elements (lists) contains the name, age & gender

person = (['Ayaan', 5, 'Male'], ['Aaradhya', 8, 'Female'])

#printing the tuple

print(person)

Output [1]: (['Ayaan', 5, 'Male'], ['Aaradhya', 8, 'Female'])

#printing the location of the object created in the memory address in hexadecimal format

print(hex(id(person)))

Output [2]: 0x1691ef47f88

#Changing the age for the 1st element. Selecting 1st element of tuple by using indexing [0] then 2nd element of the list by using indexing [1] and assigning a new value for age as 4

person[0][1] = 4

#printing the updated tuple

print(person)

Output [3]: (['Ayaan', 4, 'Male'], ['Aaradhya', 8, 'Female'])

#printing the location of the object created in the memory address in hexadecimal format

print(hex(id(person)))

Output [4]: 0x1691ef47f88

In the above code, you can see that the object ‘person’ is immutable since it is a type of tuple. However, it has two lists as it’s elements, and we can change the state of lists (lists being mutable). So, here we did not change the object reference inside the Tuple, but the referenced object was mutated.

Also Read: Real-Time Object Detection Using TensorFlow

Same way, let’s explore how it will behave if you have a mutable object which contains an immutable object? Let us again use a code to understand the behaviour–

#creating a list (mutable object) which contains tuples(immutable) as it’s elements

list1 = [(1, 2, 3), (4, 5, 6)]

#printing the list

print(list1)

Output [1]: [(1, 2, 3), (4, 5, 6)]

#printing the location of the object created in the memory address in hexadecimal format

print(hex(id(list1)))

Output [2]: 0x1691d5b13c8

#changing object reference at index 0

list1[0] = (7, 8, 9)

#printing the list

Output [3]: [(7, 8, 9), (4, 5, 6)]

#printing the location of the object created in the memory address in hexadecimal format

print(hex(id(list1)))

Output [4]: 0x1691d5b13c8

As an individual, it completely depends upon you and your requirements as to what kind of data structure you would like to create with a combination of mutable & immutable objects. I hope that this information will help you while deciding the type of object you would like to select going forward.

Before I end our discussion on IMMUTABILITY, allow me to use the word ‘CAVITE’ when we discuss the String and Integers. There is an exception, and you may see some surprising results while checking the truthiness for immutability. For instance:
#creating an object of integer type with value 10 and reference variable name ‘x’

x = 10

#printing the value of ‘x’

print(x)

Output [1]: 10

#Printing the location of the object created in the memory address in hexadecimal format

print(hex(id(x)))

Output [2]: 0x538fb560

#creating an object of integer type with value 10 and reference variable name ‘y’

y = 10

#printing the value of ‘y’

print(y)

Output [3]: 10

#Printing the location of the object created in the memory address in hexadecimal format

print(hex(id(y)))

Output [4]: 0x538fb560

As per our discussion and understanding, so far, the memory address for x & y should have been different, since, 10 is an instance of Integer class which is immutable. However, as shown in the above code, it has the same memory address. This is not something that we expected. It seems that what we have understood and discussed, has an exception as well.

Quick check – Python Data Structures

Immutability of Tuple

Tuples are immutable and hence cannot have any changes in them once they are created in Python. This is because they support the same sequence operations as strings. We all know that strings are immutable. The index operator will select an element from a tuple just like in a string. Hence, they are immutable.

Exceptions in immutability

Like all, there are exceptions in the immutability in python too. Not all immutable objects are really mutable. This will lead to a lot of doubts in your mind. Let us just take an example to understand this.

Consider a tuple ‘tup’.

Now, if we consider tuple tup = (‘GreatLearning’,[4,3,1,2]) ;

We see that the tuple has elements of different data types. The first element here is a string which as we all know is immutable in nature. The second element is a list which we all know is mutable. Now, we all know that the tuple itself is an immutable data type. It cannot change its contents. But, the list inside it can change its contents. So, the value of the Immutable objects cannot be changed but its constituent objects can. change its value.

FAQs

1. Difference between mutable vs immutable in Python?

Mutable Object	Immutable Object
State of the object can be modified after it is created.	State of the object can’t be modified once it is created.
They are not thread safe.	They are thread safe
Mutable classes are not final.	It is important to make the class final before creating an immutable object.

2. What are the mutable and immutable data types in Python?

Some mutable data types in Python are:

list, dictionary, set, user-defined classes.

Some immutable data types are:

int, float, decimal, bool, string, tuple, range.

3. Are lists mutable in Python?

Lists in Python are mutable data types as the elements of the list can be modified, individual elements can be replaced, and the order of elements can be changed even after the list has been created.
(Examples related to lists have been discussed earlier in this blog.)

4. Why are tuples called immutable types?

Tuple and list data structures are very similar, but one big difference between the data types is that lists are mutable, whereas tuples are immutable. The reason for the tuple’s immutability is that once the elements are added to the tuple and the tuple has been created; it remains unchanged.

A programmer would always prefer building a code that can be reused instead of making the whole data object again. Still, even though tuples are immutable, like lists, they can contain any Python object, including mutable objects.

5. Are sets mutable in Python?

A set is an iterable unordered collection of data type which can be used to perform mathematical operations (like union, intersection, difference etc.). Every element in a set is unique and immutable, i.e. no duplicate values should be there, and the values can’t be changed. However, we can add or remove items from the set as the set itself is mutable.

6. Are strings mutable in Python?

Strings are not mutable in Python. Strings are a immutable data types which means that its value cannot be updated.

Join Great Learning Academy’s free online courses and upgrade your skills today.

Original article source at: https://www.mygreatlearning.com

#python

Ramiro Daugherty

1672928580

How to Bash Read Command

Bash has no built-in function to take the user’s input from the terminal. The read command of Bash is used to take the user’s input from the terminal. This command has different options to take an input from the user in different ways. Multiple inputs can be taken using the single read command. Different ways of using this command in the Bash script are described in this tutorial.

Syntax

read [options] [var1, var2, var3…]

The read command can be used without any argument or option. Many types of options can be used with this command to take the input of the particular data type. It can take more input from the user by defining the multiple variables with this command.

Some Useful Options of the Read Command

Some options of the read command require an additional parameter to use. The most commonly used options of the read command are mentioned in the following:

Option	Purpose
-d <delimiter>	It is used to take the input until the delimiter value is provided.
-n <number>	It is used to take the input of a particular number of characters from the terminal and stop taking the input earlier based on the delimiter.
-N <number>	It is used to take the input of the particular number of characters from the terminal, ignoring the delimiter.
-p <prompt>	It is used to print the output of the prompt message before taking the input.
-s	It is used to take the input without an echo. This option is mainly used to take the input for the password input.
-a	It is used to take the input for the indexed array.
-t <time>	It is used to set a time limit for taking the input.
-u <file descriptor>	It is used to take the input from the file.
-r	It is used to disable the backslashes.

Different Examples of the Read Command

The uses of read command with different options are shown in this part of this tutorial.

Example 1: Using Read Command without Any Option and variable

Create a Bash file with the following script that takes the input from the terminal using the read command without any option and variable. If no variable is used with the read command, the input value is stored in the $REPLY variable. The value of this variable is printed later after taking the input.

#!/bin/bash  
#Print the prompt message
echo "Enter your favorite color: "  
#Take the input
read  
#Print the input value
echo "Your favorite color is $REPLY"

Output:

The following output appears if the “Blue” value is taken as an input:

Example 2: Using Read Command with a Variable

Create a Bash file with the following script that takes the input from the terminal using the read command with a variable. The method of taking the single or multiple variables using a read command is shown in this example. The values of all variables are printed later.

#!/bin/bash  
#Print the prompt message
echo "Enter the product name: "  
#Take the input with a single variable
read item

#Print the prompt message
echo "Enter the color variations of the product: "  
#Take three input values in three variables
read color1 color2 color3

#Print the input value
echo "The product name is $item."  
#Print the input values
echo "Available colors are $color1, $color2, and $color3."

Output:

The following output appears after taking a single input first and three inputs later:

Example 3: Using Read Command with -p Option

Create a Bash file with the following script that takes the input from the terminal using the read command with a variable and the -p option. The input value is printed later.

#!/bin/bash  
#Take the input with the prompt message
read -p "Enter the book name: " book
#Print the input value
echo "Book name: $book"

Output:

The following output appears after taking the input:

Example 4: Using Read Command with -s Option

Create a Bash file with the following script that takes the input from the terminal using the read command with a variable and the -s option. The input value of the password will not be displayed for the -s option. The input values are checked later for authentication. A success or failure message is also printed.

#!/bin/bash  
#Take the input with the prompt message
read -p "Enter your email: " email
#Take the secret input with the prompt message
read -sp "Enter your password: " password

#Add newline
echo ""

#Check the email and password for authentication
if [[ $email == "admin@example.com" && $password == "secret" ]]
then
   #Print the success message
   echo "Authenticated."
else
   #Print the failure message
   echo "Not authenticated."
fi

Output:

The following output appears after taking the valid and invalid input values:

Example 5: Using Read Command with -a Option

Create a Bash file with the following script that takes the input from the terminal using the read command with a variable and the -a option. The array values are printed later after taking the input values from the terminal.

#!/bin/bash  
echo "Enter the country names: "  
#Take multiple inputs using an array  
read -a countries

echo "Country names are:"
#Read the array values
for country in ${countries[@]}
do
    echo $country
done

Output:

The following output appears after taking the array values:

Example 6: Using Read Command with -n Option

Create a Bash file with the following script that takes the input from the terminal using the read command with a variable and the -n option.

#!/bin/bash  
#Print the prompt message
echo "Enter the product code: "  
#Take the input of five characters
read -n 5 code
#Add newline
echo ""
#Print the input value
echo "The product code is $code"

Output:

The following output appears if the “78342” value is taken as input:

Example 7: Using Read Command with -t Option

Create a Bash file with the following script that takes the input from the terminal using the read command with a variable and the -t option.

#!/bin/bash  
#Print the prompt message
echo -n "Write the result of 10-6: "  
#Take the input of five characters
read -t 3 answer

#Check the input value
if [[ $answer == "4" ]]
then
   echo "Correct answer."
else
   echo "Incorrect answer."
fi

Output:

The following output appears after taking the correct and incorrect input values:

Conclusion

The uses of some useful options of the read command are explained in this tutorial using multiple examples to know the basic uses of the read command.

Original article source at: https://linuxhint.com/

#bash #command

Chloe Butler

1667425440

Pdf2gerb: Perl Script Converts PDF Files to Gerber format

pdf2gerb

Perl script converts PDF files to Gerber format

Pdf2Gerb generates Gerber 274X photoplotting and Excellon drill files from PDFs of a PCB. Up to three PDFs are used: the top copper layer, the bottom copper layer (for 2-sided PCBs), and an optional silk screen layer. The PDFs can be created directly from any PDF drawing software, or a PDF print driver can be used to capture the Print output if the drawing software does not directly support output to PDF.

The general workflow is as follows:

Design the PCB using your favorite CAD or drawing software.
Print the top and bottom copper and top silk screen layers to a PDF file.
Run Pdf2Gerb on the PDFs to create Gerber and Excellon files.
Use a Gerber viewer to double-check the output against the original PCB design.
Make adjustments as needed.
Submit the files to a PCB manufacturer.

Please note that Pdf2Gerb does NOT perform DRC (Design Rule Checks), as these will vary according to individual PCB manufacturer conventions and capabilities. Also note that Pdf2Gerb is not perfect, so the output files must always be checked before submitting them. As of version 1.6, Pdf2Gerb supports most PCB elements, such as round and square pads, round holes, traces, SMD pads, ground planes, no-fill areas, and panelization. However, because it interprets the graphical output of a Print function, there are limitations in what it can recognize (or there may be bugs).

See docs/Pdf2Gerb.pdf for install/setup, config, usage, and other info.

pdf2gerb_cfg.pm

#Pdf2Gerb config settings:
#Put this file in same folder/directory as pdf2gerb.pl itself (global settings),
#or copy to another folder/directory with PDFs if you want PCB-specific settings.
#There is only one user of this file, so we don't need a custom package or namespace.
#NOTE: all constants defined in here will be added to main namespace.
#package pdf2gerb_cfg;

use strict; #trap undef vars (easier debug)
use warnings; #other useful info (easier debug)


##############################################################################################
#configurable settings:
#change values here instead of in main pfg2gerb.pl file

use constant WANT_COLORS => ($^O !~ m/Win/); #ANSI colors no worky on Windows? this must be set < first DebugPrint() call

#just a little warning; set realistic expectations:
#DebugPrint("${\(CYAN)}Pdf2Gerb.pl ${\(VERSION)}, $^O O/S\n${\(YELLOW)}${\(BOLD)}${\(ITALIC)}This is EXPERIMENTAL software.  \nGerber files MAY CONTAIN ERRORS.  Please CHECK them before fabrication!${\(RESET)}", 0); #if WANT_DEBUG

use constant METRIC => FALSE; #set to TRUE for metric units (only affect final numbers in output files, not internal arithmetic)
use constant APERTURE_LIMIT => 0; #34; #max #apertures to use; generate warnings if too many apertures are used (0 to not check)
use constant DRILL_FMT => '2.4'; #'2.3'; #'2.4' is the default for PCB fab; change to '2.3' for CNC

use constant WANT_DEBUG => 0; #10; #level of debug wanted; higher == more, lower == less, 0 == none
use constant GERBER_DEBUG => 0; #level of debug to include in Gerber file; DON'T USE FOR FABRICATION
use constant WANT_STREAMS => FALSE; #TRUE; #save decompressed streams to files (for debug)
use constant WANT_ALLINPUT => FALSE; #TRUE; #save entire input stream (for debug ONLY)

#DebugPrint(sprintf("${\(CYAN)}DEBUG: stdout %d, gerber %d, want streams? %d, all input? %d, O/S: $^O, Perl: $]${\(RESET)}\n", WANT_DEBUG, GERBER_DEBUG, WANT_STREAMS, WANT_ALLINPUT), 1);
#DebugPrint(sprintf("max int = %d, min int = %d\n", MAXINT, MININT), 1); 

#define standard trace and pad sizes to reduce scaling or PDF rendering errors:
#This avoids weird aperture settings and replaces them with more standardized values.
#(I'm not sure how photoplotters handle strange sizes).
#Fewer choices here gives more accurate mapping in the final Gerber files.
#units are in inches
use constant TOOL_SIZES => #add more as desired
(
#round or square pads (> 0) and drills (< 0):
    .010, -.001,  #tiny pads for SMD; dummy drill size (too small for practical use, but needed so StandardTool will use this entry)
    .031, -.014,  #used for vias
    .041, -.020,  #smallest non-filled plated hole
    .051, -.025,
    .056, -.029,  #useful for IC pins
    .070, -.033,
    .075, -.040,  #heavier leads
#    .090, -.043,  #NOTE: 600 dpi is not high enough resolution to reliably distinguish between .043" and .046", so choose 1 of the 2 here
    .100, -.046,
    .115, -.052,
    .130, -.061,
    .140, -.067,
    .150, -.079,
    .175, -.088,
    .190, -.093,
    .200, -.100,
    .220, -.110,
    .160, -.125,  #useful for mounting holes
#some additional pad sizes without holes (repeat a previous hole size if you just want the pad size):
    .090, -.040,  #want a .090 pad option, but use dummy hole size
    .065, -.040, #.065 x .065 rect pad
    .035, -.040, #.035 x .065 rect pad
#traces:
    .001,  #too thin for real traces; use only for board outlines
    .006,  #minimum real trace width; mainly used for text
    .008,  #mainly used for mid-sized text, not traces
    .010,  #minimum recommended trace width for low-current signals
    .012,
    .015,  #moderate low-voltage current
    .020,  #heavier trace for power, ground (even if a lighter one is adequate)
    .025,
    .030,  #heavy-current traces; be careful with these ones!
    .040,
    .050,
    .060,
    .080,
    .100,
    .120,
);
#Areas larger than the values below will be filled with parallel lines:
#This cuts down on the number of aperture sizes used.
#Set to 0 to always use an aperture or drill, regardless of size.
use constant { MAX_APERTURE => max((TOOL_SIZES)) + .004, MAX_DRILL => -min((TOOL_SIZES)) + .004 }; #max aperture and drill sizes (plus a little tolerance)
#DebugPrint(sprintf("using %d standard tool sizes: %s, max aper %.3f, max drill %.3f\n", scalar((TOOL_SIZES)), join(", ", (TOOL_SIZES)), MAX_APERTURE, MAX_DRILL), 1);

#NOTE: Compare the PDF to the original CAD file to check the accuracy of the PDF rendering and parsing!
#for example, the CAD software I used generated the following circles for holes:
#CAD hole size:   parsed PDF diameter:      error:
#  .014                .016                +.002
#  .020                .02267              +.00267
#  .025                .026                +.001
#  .029                .03167              +.00267
#  .033                .036                +.003
#  .040                .04267              +.00267
#This was usually ~ .002" - .003" too big compared to the hole as displayed in the CAD software.
#To compensate for PDF rendering errors (either during CAD Print function or PDF parsing logic), adjust the values below as needed.
#units are pixels; for example, a value of 2.4 at 600 dpi = .0004 inch, 2 at 600 dpi = .0033"
use constant
{
    HOLE_ADJUST => -0.004 * 600, #-2.6, #holes seemed to be slightly oversized (by .002" - .004"), so shrink them a little
    RNDPAD_ADJUST => -0.003 * 600, #-2, #-2.4, #round pads seemed to be slightly oversized, so shrink them a little
    SQRPAD_ADJUST => +0.001 * 600, #+.5, #square pads are sometimes too small by .00067, so bump them up a little
    RECTPAD_ADJUST => 0, #(pixels) rectangular pads seem to be okay? (not tested much)
    TRACE_ADJUST => 0, #(pixels) traces seemed to be okay?
    REDUCE_TOLERANCE => .001, #(inches) allow this much variation when reducing circles and rects
};

#Also, my CAD's Print function or the PDF print driver I used was a little off for circles, so define some additional adjustment values here:
#Values are added to X/Y coordinates; units are pixels; for example, a value of 1 at 600 dpi would be ~= .002 inch
use constant
{
    CIRCLE_ADJUST_MINX => 0,
    CIRCLE_ADJUST_MINY => -0.001 * 600, #-1, #circles were a little too high, so nudge them a little lower
    CIRCLE_ADJUST_MAXX => +0.001 * 600, #+1, #circles were a little too far to the left, so nudge them a little to the right
    CIRCLE_ADJUST_MAXY => 0,
    SUBST_CIRCLE_CLIPRECT => FALSE, #generate circle and substitute for clip rects (to compensate for the way some CAD software draws circles)
    WANT_CLIPRECT => TRUE, #FALSE, #AI doesn't need clip rect at all? should be on normally?
    RECT_COMPLETION => FALSE, #TRUE, #fill in 4th side of rect when 3 sides found
};

#allow .012 clearance around pads for solder mask:
#This value effectively adjusts pad sizes in the TOOL_SIZES list above (only for solder mask layers).
use constant SOLDER_MARGIN => +.012; #units are inches

#line join/cap styles:
use constant
{
    CAP_NONE => 0, #butt (none); line is exact length
    CAP_ROUND => 1, #round cap/join; line overhangs by a semi-circle at either end
    CAP_SQUARE => 2, #square cap/join; line overhangs by a half square on either end
    CAP_OVERRIDE => FALSE, #cap style overrides drawing logic
};
    
#number of elements in each shape type:
use constant
{
    RECT_SHAPELEN => 6, #x0, y0, x1, y1, count, "rect" (start, end corners)
    LINE_SHAPELEN => 6, #x0, y0, x1, y1, count, "line" (line seg)
    CURVE_SHAPELEN => 10, #xstart, ystart, x0, y0, x1, y1, xend, yend, count, "curve" (bezier 2 points)
    CIRCLE_SHAPELEN => 5, #x, y, 5, count, "circle" (center + radius)
};
#const my %SHAPELEN =
#Readonly my %SHAPELEN =>
our %SHAPELEN =
(
    rect => RECT_SHAPELEN,
    line => LINE_SHAPELEN,
    curve => CURVE_SHAPELEN,
    circle => CIRCLE_SHAPELEN,
);

#panelization:
#This will repeat the entire body the number of times indicated along the X or Y axes (files grow accordingly).
#Display elements that overhang PCB boundary can be squashed or left as-is (typically text or other silk screen markings).
#Set "overhangs" TRUE to allow overhangs, FALSE to truncate them.
#xpad and ypad allow margins to be added around outer edge of panelized PCB.
use constant PANELIZE => {'x' => 1, 'y' => 1, 'xpad' => 0, 'ypad' => 0, 'overhangs' => TRUE}; #number of times to repeat in X and Y directions

# Set this to 1 if you need TurboCAD support.
#$turboCAD = FALSE; #is this still needed as an option?

#CIRCAD pad generation uses an appropriate aperture, then moves it (stroke) "a little" - we use this to find pads and distinguish them from PCB holes. 
use constant PAD_STROKE => 0.3; #0.0005 * 600; #units are pixels
#convert very short traces to pads or holes:
use constant TRACE_MINLEN => .001; #units are inches
#use constant ALWAYS_XY => TRUE; #FALSE; #force XY even if X or Y doesn't change; NOTE: needs to be TRUE for all pads to show in FlatCAM and ViewPlot
use constant REMOVE_POLARITY => FALSE; #TRUE; #set to remove subtractive (negative) polarity; NOTE: must be FALSE for ground planes

#PDF uses "points", each point = 1/72 inch
#combined with a PDF scale factor of .12, this gives 600 dpi resolution (1/72 * .12 = 600 dpi)
use constant INCHES_PER_POINT => 1/72; #0.0138888889; #multiply point-size by this to get inches

# The precision used when computing a bezier curve. Higher numbers are more precise but slower (and generate larger files).
#$bezierPrecision = 100;
use constant BEZIER_PRECISION => 36; #100; #use const; reduced for faster rendering (mainly used for silk screen and thermal pads)

# Ground planes and silk screen or larger copper rectangles or circles are filled line-by-line using this resolution.
use constant FILL_WIDTH => .01; #fill at most 0.01 inch at a time

# The max number of characters to read into memory
use constant MAX_BYTES => 10 * M; #bumped up to 10 MB, use const

use constant DUP_DRILL1 => TRUE; #FALSE; #kludge: ViewPlot doesn't load drill files that are too small so duplicate first tool

my $runtime = time(); #Time::HiRes::gettimeofday(); #measure my execution time

print STDERR "Loaded config settings from '${\(__FILE__)}'.\n";
1; #last value must be truthful to indicate successful load


#############################################################################################
#junk/experiment:

#use Package::Constants;
#use Exporter qw(import); #https://perldoc.perl.org/Exporter.html

#my $caller = "pdf2gerb::";

#sub cfg
#{
#    my $proto = shift;
#    my $class = ref($proto) || $proto;
#    my $settings =
#    {
#        $WANT_DEBUG => 990, #10; #level of debug wanted; higher == more, lower == less, 0 == none
#    };
#    bless($settings, $class);
#    return $settings;
#}

#use constant HELLO => "hi there2"; #"main::HELLO" => "hi there";
#use constant GOODBYE => 14; #"main::GOODBYE" => 12;

#print STDERR "read cfg file\n";

#our @EXPORT_OK = Package::Constants->list(__PACKAGE__); #https://www.perlmonks.org/?node_id=1072691; NOTE: "_OK" skips short/common names

#print STDERR scalar(@EXPORT_OK) . " consts exported:\n";
#foreach(@EXPORT_OK) { print STDERR "$_\n"; }
#my $val = main::thing("xyz");
#print STDERR "caller gave me $val\n";
#foreach my $arg (@ARGV) { print STDERR "arg $arg\n"; }

Download Details:

Author: swannman
Source Code: https://github.com/swannman/pdf2gerb

License: GPL-3.0 license

#perl

Hoang Kim

1635843812

Chatbot AI hội thoại với Máy biến áp được đào tạo trước bằng Python

Tìm hiểu cách sử dụng thư viện máy biến áp Huggingface để tạo phản hồi hội thoại bằng mô hình DialoGPT được đào tạo trước bằng Python.

Chatbots đã trở nên phổ biến trong những năm gần đây và khi mối quan tâm ngày càng tăng trong việc sử dụng chatbots cho kinh doanh, các nhà nghiên cứu cũng đã làm rất tốt trong việc phát triển các chatbot AI đàm thoại.

Trong hướng dẫn này, chúng tôi sẽ sử dụng thư viện máy biến áp Huggingface để sử dụng mô hình DialoGPT đã được đào tạo trước để tạo phản hồi hội thoại.

DialoGPT là một mô hình tạo phản hồi hội thoại thần kinh có thể điều chỉnh quy mô lớn được đào tạo trên 147 triệu cuộc hội thoại được trích xuất từ Reddit và điều tốt là bạn có thể tinh chỉnh nó với bộ dữ liệu của mình để đạt được hiệu suất tốt hơn so với đào tạo từ đầu.

Để bắt đầu, hãy cài đặt máy biến áp :

$ pip3 install transformers

Mở tệp hoặc sổ ghi chép Python mới và thực hiện như sau:

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

# model_name = "microsoft/DialoGPT-large"
model_name = "microsoft/DialoGPT-medium"
# model_name = "microsoft/DialoGPT-small"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)

Có ba phiên bản của DialoGPT; nhỏ, vừa và lớn. Tất nhiên, càng lớn càng tốt, nhưng nếu bạn chạy điều này trên máy của mình, tôi nghĩ nhỏ hoặc trung bình phù hợp với bộ nhớ của bạn mà không có vấn đề gì. Bạn cũng có thể sử dụng Google Colab để thử cái lớn.

Tạo phản hồi bằng Tìm kiếm tham lam

Trong phần này, chúng tôi sẽ sử dụng thuật toán tìm kiếm tham lam để tạo phản hồi. Đó là, chúng tôi chọn phản hồi chatbot có xác suất cao nhất được chọn trên mỗi bước thời gian.

Hãy tạo mã để trò chuyện với AI của chúng tôi bằng cách sử dụng tìm kiếm tham lam:

# chatting 5 times with greedy search
for step in range(5):
    # take user input
    text = input(">> You:")
    # encode the input and add end of string token
    input_ids = tokenizer.encode(text + tokenizer.eos_token, return_tensors="pt")
    # concatenate new user input with chat history (if there is)
    bot_input_ids = torch.cat([chat_history_ids, input_ids], dim=-1) if step > 0 else input_ids
    # generate a bot response
    chat_history_ids = model.generate(
        bot_input_ids,
        max_length=1000,
        pad_token_id=tokenizer.eos_token_id,
    )
    #print the output
    output = tokenizer.decode(chat_history_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True)
    print(f"DialoGPT: {output}")

Hãy giải thích cốt lõi của mã này:

Đầu tiên chúng tôi lấy thông tin đầu vào từ người dùng để trò chuyện.
Chúng tôi mã hóa văn bản input_idsbằng cách sử dụng trình mã hóa DialoGPT, chúng tôi cũng nối vào cuối chuỗi mã thông báo và trả về nó dưới dạng bộ căng Pytorch.
Nếu đây là lần đầu tiên trò chuyện với bot, thì chúng tôi sẽ trực tiếp cung cấp input_idscho mô hình của mình trong một thế hệ. Nếu không, chúng tôi nối lịch sử trò chuyện bằng cách nối với sự trợ giúp của torch.cat()phương thức.
Sau đó, chúng tôi sử dụng model.generate()phương pháp tạo phản hồi chatbot.
Cuối cùng, như sản lượng trở lại là một chuỗi tokenized quá, chúng tôi giải mã trình tự sử dụng tokenizer.decode()và thiết lập skip_special_tokensđể Trueđảm bảo chúng tôi không thấy bất kỳ đặc biệt gây phiền nhiễu mã thông báo như <|endoftext|>. Ngoài ra, vì mô hình trả về toàn bộ chuỗi, chúng tôi bỏ qua lịch sử trò chuyện trước đó và chỉ in câu trả lời chatbot mới được tạo.

Dưới đây là một cuộc thảo luận mẫu với bot:

>> You:How can you be rich so quickly?
DialoGPT: I'm not rich, I'm just a rich man.
>> You:Ok, you don't make sense, how old are you ?
DialoGPT: I'm not a rich man, I'm just a rich man.
>> You:How can you be happy ?
DialoGPT: I'm not happy, I'm just a rich man.
>> You:But they say money buys happiness!
DialoGPT: Money buys happiness, but not happiness itself.
>> You:You don't make sense again!
DialoGPT: I'm not happy, I'm just a rich man.

Bạn thấy mô hình lặp lại rất nhiều phản hồi, vì đây là xác suất cao nhất và nó luôn chọn nó.

Theo mặc định, model.generate()sử dụng thuật toán tìm kiếm tham lam khi không có tham số nào khác được đặt, trong các phần tiếp theo, chúng tôi sẽ thêm một số đối số vào phương thức này để xem liệu chúng tôi có thể cải thiện việc tạo không.

Tạo phản hồi với Tìm kiếm chùm

Tìm kiếm theo chùm cho phép chúng tôi giảm nguy cơ bỏ lỡ các chuỗi có xác suất cao bằng cách giữ lại num_beamscác giả thuyết có khả năng xảy ra nhất ở mỗi bước thời gian và sau đó lấy các chuỗi có xác suất cao nhất tổng thể, đoạn mã dưới đây sẽ tạo ra các phản hồi của chatbot với tìm kiếm theo chùm:

# chatting 5 times with beam search
for step in range(5):
    # take user input
    text = input(">> You:")
    # encode the input and add end of string token
    input_ids = tokenizer.encode(text + tokenizer.eos_token, return_tensors="pt")
    # concatenate new user input with chat history (if there is)
    bot_input_ids = torch.cat([chat_history_ids, input_ids], dim=-1) if step > 0 else input_ids
    # generate a bot response
    chat_history_ids = model.generate(
        bot_input_ids,
        max_length=1000,
        num_beams=3,
        early_stopping=True,
        pad_token_id=tokenizer.eos_token_id
    )
    #print the output
    output = tokenizer.decode(chat_history_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True)
    print(f"DialoGPT: {output}")

Khi đặt num_beamsthành 3trong model.generate()phương thức, chúng tôi sẽ chọn 3 từ ở mỗi bước và phát triển chúng để tìm xác suất tổng thể cao nhất của chuỗi, đặt num_beamsthành 1 cũng giống như tìm kiếm tham lam.

Dưới đây là một cuộc thảo luận mẫu với chatbot bằng cách sử dụng tìm kiếm chùm:

>> You:How can I be rich?
DialoGPT: You can't.
>> You:Why not?
DialoGPT: Because you can't.
>> You:Then how am I supposed to live well?
DialoGPT: You can't.
>> You:So basically, you want me to be a poor guy?
DialoGPT: No, I want you to be a rich guy.
>> You:Ok how to be rich, then?
DialoGPT: I don't know.

Tạo phản hồi bằng lấy mẫu

Trong các phần trước, chúng tôi đã sử dụng tìm kiếm chùm và tham lam để tạo ra chuỗi xác suất cao nhất. Giờ đây, điều đó thật tuyệt vời cho các tác vụ như dịch máy hoặc tóm tắt văn bản trong đó kết quả đầu ra có thể dự đoán được. Tuy nhiên, nó không phải là lựa chọn tốt nhất cho thế hệ kết thúc mở như trong chatbots.

Để có một thế hệ tốt hơn, chúng tôi cần đưa ra một số ngẫu nhiên trong đó chúng tôi lấy mẫu từ một loạt các trình tự ứng viên dựa trên xác suất:

# chatting 5 times with sampling
for step in range(5):
    # take user input
    text = input(">> You:")
    # encode the input and add end of string token
    input_ids = tokenizer.encode(text + tokenizer.eos_token, return_tensors="pt")
    # concatenate new user input with chat history (if there is)
    bot_input_ids = torch.cat([chat_history_ids, input_ids], dim=-1) if step > 0 else input_ids
    # generate a bot response
    chat_history_ids = model.generate(
        bot_input_ids,
        max_length=1000,
        do_sample=True,
        top_k=0,
        pad_token_id=tokenizer.eos_token_id
    )
    #print the output
    output = tokenizer.decode(chat_history_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True)
    print(f"DialoGPT: {output}")

Lần này, chúng tôi đặt do_sampleđể Truelấy mẫu và chúng tôi đặt top_kđể 0chỉ ra rằng chúng tôi đang chọn tất cả các xác suất có thể xảy ra, sau đó chúng ta sẽ thảo luận về top_ktham số.

Đây là một cuộc trò chuyện với các thông số này:

>> You:Hello, how can I be rich?
DialoGPT: You can't
>> You:Really? then how others get rich?
DialoGPT: They work as money artists
>> You:How to be a money artist?
DialoGPT: I don't know, I don't have enough knowledge on artists
>> You:Alright, thank you!
DialoGPT: No problem, if you need tips pm me.
>> You:Great, you're my savior!
DialoGPT: Thank you for helping a stranger out like me!

Rõ ràng là có một số cải tiến. Tuy nhiên, việc lấy mẫu trên một danh sách rộng các chuỗi với xác suất thấp có thể dẫn đến việc tạo ngẫu nhiên (như bạn thấy trong câu cuối cùng).

Để cải thiện nó hơn nữa, chúng tôi có thể:

Hạ thấp mẫu temperature, điều đó giúp chúng tôi giảm khả năng chọn các từ có xác suất thấp và tăng khả năng chọn các từ có xác suất cao.
Sử dụng lấy mẫu Top-k thay vì chọn tất cả các lần xuất hiện có thể xảy ra, điều này sẽ giúp chúng tôi loại bỏ các từ có xác suất thấp để được chọn.

# chatting 5 times with Top K sampling & tweaking temperature
for step in range(5):
    # take user input
    text = input(">> You:")
    # encode the input and add end of string token
    input_ids = tokenizer.encode(text + tokenizer.eos_token, return_tensors="pt")
    # concatenate new user input with chat history (if there is)
    bot_input_ids = torch.cat([chat_history_ids, input_ids], dim=-1) if step > 0 else input_ids
    # generate a bot response
    chat_history_ids = model.generate(
        bot_input_ids,
        max_length=1000,
        do_sample=True,
        top_k=100,
        temperature=0.75,
        pad_token_id=tokenizer.eos_token_id
    )
    #print the output
    output = tokenizer.decode(chat_history_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True)
    print(f"DialoGPT: {output}")

Bây giờ, chúng ta thiết lập top_kđể 100lấy mẫu từ đỉnh 100từ được sắp xếp descendingly bởi xác suất. Chúng tôi cũng đặt temperaturethành 0.75(mặc định là 1.0) để có cơ hội chọn các từ có xác suất cao hơn, đặt nhiệt độ 0.0giống như tìm kiếm tham lam, đặt nhiệt độ thành vô cùng giống như hoàn toàn ngẫu nhiên.

Đây là một cuộc thảo luận với các thông số này:

>> You:How can I get rich?
DialoGPT: I'd also like to know this.
>> You:So you don't know it?
DialoGPT: I know it, but it doesn't matter.
>> You:Really? money matters, man!
DialoGPT: Well, money is money and I gotta get my money.
>> You:Yeah how to get them?
DialoGPT: Well, that's a question. I'd like to know more, just for curiosity's sake, to know what other people have said about how I should get rich.
>> You:Some robots like you told me you have to buy gold, silver or stocks.
DialoGPT: I think my wallet is empty.

Như bạn có thể thấy, bây giờ nó đã tốt hơn nhiều, hãy thoải mái tinh chỉnh temperaturevà top_kcác thông số và xem liệu nó có thể cải thiện nó hay không.

Lấy mẫu hạt nhân

Lấy mẫu hạt nhân hoặc lấy mẫu Top-p chọn từ các từ nhỏ nhất có thể có xác suất tích lũy vượt quá tham số pchúng tôi đặt.

Dưới đây là một ví dụ sử dụng lấy mẫu Top-p:

# chatting 5 times with nucleus sampling & tweaking temperature
for step in range(5):
    # take user input
    text = input(">> You:")
    # encode the input and add end of string token
    input_ids = tokenizer.encode(text + tokenizer.eos_token, return_tensors="pt")
    # concatenate new user input with chat history (if there is)
    bot_input_ids = torch.cat([chat_history_ids, input_ids], dim=-1) if step > 0 else input_ids
    # generate a bot response
    chat_history_ids = model.generate(
        bot_input_ids,
        max_length=1000,
        do_sample=True,
        top_p=0.95,
        top_k=0,
        temperature=0.75,
        pad_token_id=tokenizer.eos_token_id
    )
    #print the output
    output = tokenizer.decode(chat_history_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True)
    print(f"DialoGPT: {output}")

Chúng tôi đặt top_kđể 0tắt lấy mẫu Top-k, nhưng bạn có thể sử dụng cả hai phương pháp có xu hướng hoạt động tốt hơn. Đây là một cuộc trò chuyện:

>> You:How can I be rich fast?
DialoGPT: Inflation and free money
>> You:lol, but that takes time, I need fast money.
DialoGPT: Your post makes no sense. You just need a lot of money to be rich fast
>> You:Ok how to get a lot of money?
DialoGPT: Buy gold and then trade it for money.
>> You:Hmm, I got your point, thanks.
DialoGPT: No problem. You have to have money to buy gold, but its only good for trading it.
>> You:Yeah, I understand, thank you so much.
DialoGPT: Ok, let me know if you need help

Giờ đây, chatbot rõ ràng có ý nghĩa trong nhiều trường hợp.

Bây giờ, hãy thêm một số mã để tạo nhiều hơn một phản hồi chatbot và sau đó chúng tôi chọn phản hồi nào sẽ bao gồm trong đầu vào tiếp theo:

# chatting 5 times with nucleus & top-k sampling & tweaking temperature & multiple
# sentences
for step in range(5):
    # take user input
    text = input(">> You:")
    # encode the input and add end of string token
    input_ids = tokenizer.encode(text + tokenizer.eos_token, return_tensors="pt")
    # concatenate new user input with chat history (if there is)
    bot_input_ids = torch.cat([chat_history_ids, input_ids], dim=-1) if step > 0 else input_ids
    # generate a bot response
    chat_history_ids_list = model.generate(
        bot_input_ids,
        max_length=1000,
        do_sample=True,
        top_p=0.95,
        top_k=50,
        temperature=0.75,
        num_return_sequences=5,
        pad_token_id=tokenizer.eos_token_id
    )
    #print the outputs
    for i in range(len(chat_history_ids_list)):
      output = tokenizer.decode(chat_history_ids_list[i][bot_input_ids.shape[-1]:], skip_special_tokens=True)
      print(f"DialoGPT {i}: {output}")
    choice_index = int(input("Choose the response you want for the next input: "))
    chat_history_ids = torch.unsqueeze(chat_history_ids_list[choice_index], dim=0)

Tôi đã thiết lập num_return_sequencesđể 5trả lại 5 câu cùng một lúc, chúng ta phải chọn một câu sẽ có trong chuỗi tiếp theo. Dưới đây là làm thế nào nó đi:

>> You:Hello, friend.
DialoGPT 0: Hey there!
DialoGPT 1: I am your friend, and I am also your friend.
DialoGPT 2: Hello. :
DialoGPT 3: Hiya! How are you?
DialoGPT 4: Hello. I like your username.
Choose the response you want for the next input: 3
>> You:Fine, thanks, what about you?
DialoGPT 0: Good. Not much to do.
DialoGPT 1: Not much, you?
DialoGPT 2: Fine, thanks.
DialoGPT 3: I'm okay, thank you for asking.
DialoGPT 4: I'm fine too.
Choose the response you want for the next input: 2
>> You:Can you tell me how can I get rich quickly?
DialoGPT 0: Yes, I can!
DialoGPT 1: Yes, just look up how to get rich on Google.
DialoGPT 2: Just keep doing your thing. You'll get rich.
DialoGPT 3: No, I can't.
DialoGPT 4: no, i have no money
Choose the response you want for the next input: 1
>> You:Yeah, but I thought you have an answer...
DialoGPT 0: I'm getting married in the morning.
DialoGPT 1: How do you get rich?
DialoGPT 2: I think there are many good answers, but I could send you one, if you want.
DialoGPT 3: I will, thanks!
DialoGPT 4: I thought you have an answer.
Choose the response you want for the next input: 2
>> You:Yes, please send me.
DialoGPT 0: You can send me link.
DialoGPT 1: I am very interested
DialoGPT 2: I sent you a PM
DialoGPT 3: I'll send you a PM
DialoGPT 4: I am always interested in new ideas.
Choose the response you want for the next input: 2

Phần kết luận

Và bạn hiểu rồi, tôi hy vọng hướng dẫn này đã giúp bạn cách tạo văn bản trên DialoGPT và các mô hình tương tự. Để biết thêm thông tin về cách tạo văn bản, tôi thực sự khuyên bạn nên đọc hướng dẫn Cách tạo văn bản bằng Transformers .

Tôi sẽ để bạn điều chỉnh các thông số để xem liệu bạn có thể làm cho bot hoạt động tốt hơn hay không.

Ngoài ra, bạn có thể kết hợp điều này với các hướng dẫn chuyển văn bản thành giọng nói và chuyển lời nói thành văn bản để xây dựng một trợ lý ảo như Alexa , Siri , Cortana , v.v.

#python #ai #chatbot

Berta Escamilla

1635844372

Chatbot conversacional de IA con Transformers en Python

Aprenda a usar la biblioteca de transformadores Huggingface para generar respuestas conversacionales con el modelo DialoGPT previamente entrenado en Python.

Los chatbots han ganado mucha popularidad en los últimos años y, a medida que crece el interés en el uso de chatbots para empresas, los investigadores también hicieron un gran trabajo en el avance de los chatbots de IA conversacionales.

En este tutorial, usaremos la biblioteca de transformadores Huggingface para emplear el modelo DialoGPT previamente entrenado para la generación de respuestas conversacionales.

DialoGPT es un modelo de generación de respuesta conversacional neuronal sintonizable a gran escala que se entrenó en 147 millones de conversaciones extraídas de Reddit, y lo bueno es que puede ajustarlo con su conjunto de datos para lograr un mejor rendimiento que el entrenamiento desde cero.

Para comenzar, instalemos transformadores :

$ pip3 install transformers

Abra un nuevo archivo o cuaderno de Python y haga lo siguiente:

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

# model_name = "microsoft/DialoGPT-large"
model_name = "microsoft/DialoGPT-medium"
# model_name = "microsoft/DialoGPT-small"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)

Hay tres versiones de DialoGPT; pequeño, mediano y grande. Por supuesto, cuanto más grande, mejor, pero si ejecuta esto en su máquina, creo que el tamaño pequeño o mediano se adapta a su memoria sin problemas. También puede utilizar Google Colab para probar el más grande.

Generación de respuestas con búsqueda codiciosa

En esta sección, usaremos el algoritmo de búsqueda codiciosa para generar respuestas. Es decir, seleccionamos la respuesta del chatbot que tiene la mayor probabilidad de ser seleccionada en cada paso de tiempo.

Hagamos un código para chatear con nuestra IA usando una búsqueda codiciosa:

# chatting 5 times with greedy search
for step in range(5):
    # take user input
    text = input(">> You:")
    # encode the input and add end of string token
    input_ids = tokenizer.encode(text + tokenizer.eos_token, return_tensors="pt")
    # concatenate new user input with chat history (if there is)
    bot_input_ids = torch.cat([chat_history_ids, input_ids], dim=-1) if step > 0 else input_ids
    # generate a bot response
    chat_history_ids = model.generate(
        bot_input_ids,
        max_length=1000,
        pad_token_id=tokenizer.eos_token_id,
    )
    #print the output
    output = tokenizer.decode(chat_history_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True)
    print(f"DialoGPT: {output}")

Expliquemos el núcleo de este código:

Primero tomamos información del usuario para chatear.
Codificamos el texto para input_idsusar el tokenizador DialoGPT, también agregamos el final del token de cadena y lo devolvemos como un tensor de Pytorch.
Si esta es la primera vez que chateamos con el bot, alimentaremos directamente input_idsnuestro modelo durante una generación. De lo contrario, agregamos el historial de chat usando la concatenación con la ayuda del torch.cat()método.
Después de eso, usamos el model.generate()método para generar la respuesta del chatbot.
Por último, como la salida devuelta también es una secuencia tokenizada, decodificamos la secuencia usando tokenizer.decode()y configuramos skip_special_tokenspara Trueasegurarnos de que no veamos ningún token especial molesto como <|endoftext|>. Además, dado que el modelo devuelve la secuencia completa, omitimos el historial de chat anterior e imprimimos solo la respuesta del chatbot recién generada.

A continuación se muestra una discusión de muestra con el bot:

>> You:How can you be rich so quickly?
DialoGPT: I'm not rich, I'm just a rich man.
>> You:Ok, you don't make sense, how old are you ?
DialoGPT: I'm not a rich man, I'm just a rich man.
>> You:How can you be happy ?
DialoGPT: I'm not happy, I'm just a rich man.
>> You:But they say money buys happiness!
DialoGPT: Money buys happiness, but not happiness itself.
>> You:You don't make sense again!
DialoGPT: I'm not happy, I'm just a rich man.

Verá que el modelo repite muchas respuestas, ya que estas son la probabilidad más alta y lo elige cada vez.

De forma predeterminada, model.generate()utiliza un algoritmo de búsqueda codicioso cuando no se establecen otros parámetros; en las siguientes secciones, agregaremos algunos argumentos a este método para ver si podemos mejorar la generación.

Generación de respuestas con Beam Search

La búsqueda por haz nos permite reducir el riesgo de perder secuencias de alta probabilidad al mantener las num_beamshipótesis más probables en cada paso de tiempo y luego tomar las secuencias que tienen la probabilidad general más alta, el siguiente código generará respuestas de chatbot con búsqueda de haz:

# chatting 5 times with beam search
for step in range(5):
    # take user input
    text = input(">> You:")
    # encode the input and add end of string token
    input_ids = tokenizer.encode(text + tokenizer.eos_token, return_tensors="pt")
    # concatenate new user input with chat history (if there is)
    bot_input_ids = torch.cat([chat_history_ids, input_ids], dim=-1) if step > 0 else input_ids
    # generate a bot response
    chat_history_ids = model.generate(
        bot_input_ids,
        max_length=1000,
        num_beams=3,
        early_stopping=True,
        pad_token_id=tokenizer.eos_token_id
    )
    #print the output
    output = tokenizer.decode(chat_history_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True)
    print(f"DialoGPT: {output}")

Cuando se ajusta num_beamsa 3de model.generate()método, a continuación, vamos a seleccionar 3 palabras en cada paso de tiempo y desarrollarlas para encontrar la más alta probabilidad global de la secuencia, el establecimiento num_beamsde 1 es la misma que la búsqueda codiciosa.

A continuación se muestra una discusión de muestra con el chatbot usando la búsqueda de haz:

>> You:How can I be rich?
DialoGPT: You can't.
>> You:Why not?
DialoGPT: Because you can't.
>> You:Then how am I supposed to live well?
DialoGPT: You can't.
>> You:So basically, you want me to be a poor guy?
DialoGPT: No, I want you to be a rich guy.
>> You:Ok how to be rich, then?
DialoGPT: I don't know.

Generando respuestas con muestreo

En las secciones anteriores, usamos la búsqueda de rayos y codiciosos para generar la secuencia de probabilidad más alta. Eso es genial para tareas como la traducción automática o el resumen de texto donde el resultado es predecible. Sin embargo, no es la mejor opción para una generación abierta como en los chatbots.

Para una mejor generación, necesitamos introducir algo de aleatoriedad donde muestreamos de una amplia gama de secuencias candidatas basadas en probabilidades:

# chatting 5 times with sampling
for step in range(5):
    # take user input
    text = input(">> You:")
    # encode the input and add end of string token
    input_ids = tokenizer.encode(text + tokenizer.eos_token, return_tensors="pt")
    # concatenate new user input with chat history (if there is)
    bot_input_ids = torch.cat([chat_history_ids, input_ids], dim=-1) if step > 0 else input_ids
    # generate a bot response
    chat_history_ids = model.generate(
        bot_input_ids,
        max_length=1000,
        do_sample=True,
        top_k=0,
        pad_token_id=tokenizer.eos_token_id
    )
    #print the output
    output = tokenizer.decode(chat_history_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True)
    print(f"DialoGPT: {output}")

Esta vez, establece do_sampleque Truepara el muestreo, y nos pusimos top_ka 0lo que indica que estamos seleccionando todas las probabilidades posibles, vamos a discutir más adelante top_kparámetro.

Aquí hay un chat con estos parámetros:

>> You:Hello, how can I be rich?
DialoGPT: You can't
>> You:Really? then how others get rich?
DialoGPT: They work as money artists
>> You:How to be a money artist?
DialoGPT: I don't know, I don't have enough knowledge on artists
>> You:Alright, thank you!
DialoGPT: No problem, if you need tips pm me.
>> You:Great, you're my savior!
DialoGPT: Thank you for helping a stranger out like me!

Claramente hay algunas mejoras. Sin embargo, el muestreo en una amplia lista de secuencias con bajas probabilidades puede conducir a una generación aleatoria (como se ve en la última oración).

Para mejorarlo aún más, podemos:

temperatureDisminuir el muestreo , eso nos ayuda a disminuir la probabilidad de elegir palabras de baja probabilidad y aumentar la probabilidad de elegir palabras de alta probabilidad.
Utilice el muestreo de Top-k en lugar de seleccionar todas las ocurrencias probables, esto nos ayudará a descartar palabras de baja probabilidad para que no sean seleccionadas.

# chatting 5 times with Top K sampling & tweaking temperature
for step in range(5):
    # take user input
    text = input(">> You:")
    # encode the input and add end of string token
    input_ids = tokenizer.encode(text + tokenizer.eos_token, return_tensors="pt")
    # concatenate new user input with chat history (if there is)
    bot_input_ids = torch.cat([chat_history_ids, input_ids], dim=-1) if step > 0 else input_ids
    # generate a bot response
    chat_history_ids = model.generate(
        bot_input_ids,
        max_length=1000,
        do_sample=True,
        top_k=100,
        temperature=0.75,
        pad_token_id=tokenizer.eos_token_id
    )
    #print the output
    output = tokenizer.decode(chat_history_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True)
    print(f"DialoGPT: {output}")

Ahora, configuramos top_kpara 100muestrear las 100palabras principales ordenadas de forma descendente por probabilidad. También establecemos temperatureen 0.75(el valor predeterminado es 1.0) para brindar una mayor probabilidad de elegir palabras de alta probabilidad, establecer la temperatura en 0.0es lo mismo que la búsqueda codiciosa, establecerla en infinito es lo mismo que completamente aleatorio.

Aquí hay una discusión con estos parámetros:

>> You:How can I get rich?
DialoGPT: I'd also like to know this.
>> You:So you don't know it?
DialoGPT: I know it, but it doesn't matter.
>> You:Really? money matters, man!
DialoGPT: Well, money is money and I gotta get my money.
>> You:Yeah how to get them?
DialoGPT: Well, that's a question. I'd like to know more, just for curiosity's sake, to know what other people have said about how I should get rich.
>> You:Some robots like you told me you have to buy gold, silver or stocks.
DialoGPT: I think my wallet is empty.

Como se puede ver, es mucho mejor ahora, no dude en modificar temperaturey top_kparámetros y ver si puede mejorarlo.

Muestreo de núcleos

El muestreo de núcleo o muestreo de Top-p elige entre las palabras más pequeñas posibles cuya probabilidad acumulada excede el parámetro pque establecimos.

A continuación se muestra un ejemplo con el muestreo de Top-p:

# chatting 5 times with nucleus sampling & tweaking temperature
for step in range(5):
    # take user input
    text = input(">> You:")
    # encode the input and add end of string token
    input_ids = tokenizer.encode(text + tokenizer.eos_token, return_tensors="pt")
    # concatenate new user input with chat history (if there is)
    bot_input_ids = torch.cat([chat_history_ids, input_ids], dim=-1) if step > 0 else input_ids
    # generate a bot response
    chat_history_ids = model.generate(
        bot_input_ids,
        max_length=1000,
        do_sample=True,
        top_p=0.95,
        top_k=0,
        temperature=0.75,
        pad_token_id=tokenizer.eos_token_id
    )
    #print the output
    output = tokenizer.decode(chat_history_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True)
    print(f"DialoGPT: {output}")

Configuramos top_kpara 0deshabilitar el muestreo de Top-k, pero puede usar ambos métodos, lo que tiende a funcionar mejor. Aquí hay un chat:

>> You:How can I be rich fast?
DialoGPT: Inflation and free money
>> You:lol, but that takes time, I need fast money.
DialoGPT: Your post makes no sense. You just need a lot of money to be rich fast
>> You:Ok how to get a lot of money?
DialoGPT: Buy gold and then trade it for money.
>> You:Hmm, I got your point, thanks.
DialoGPT: No problem. You have to have money to buy gold, but its only good for trading it.
>> You:Yeah, I understand, thank you so much.
DialoGPT: Ok, let me know if you need help

Ahora el chatbot claramente tiene sentido en muchos casos.

Ahora agreguemos algo de código para generar más de una respuesta de chatbot, y luego elegimos qué respuesta incluir en la siguiente entrada:

# chatting 5 times with nucleus & top-k sampling & tweaking temperature & multiple
# sentences
for step in range(5):
    # take user input
    text = input(">> You:")
    # encode the input and add end of string token
    input_ids = tokenizer.encode(text + tokenizer.eos_token, return_tensors="pt")
    # concatenate new user input with chat history (if there is)
    bot_input_ids = torch.cat([chat_history_ids, input_ids], dim=-1) if step > 0 else input_ids
    # generate a bot response
    chat_history_ids_list = model.generate(
        bot_input_ids,
        max_length=1000,
        do_sample=True,
        top_p=0.95,
        top_k=50,
        temperature=0.75,
        num_return_sequences=5,
        pad_token_id=tokenizer.eos_token_id
    )
    #print the outputs
    for i in range(len(chat_history_ids_list)):
      output = tokenizer.decode(chat_history_ids_list[i][bot_input_ids.shape[-1]:], skip_special_tokens=True)
      print(f"DialoGPT {i}: {output}")
    choice_index = int(input("Choose the response you want for the next input: "))
    chat_history_ids = torch.unsqueeze(chat_history_ids_list[choice_index], dim=0)

He configurado num_return_sequencespara 5devolver 5 oraciones a la vez, tenemos que elegir la que se incluirá en la siguiente secuencia. Así es como fue:

>> You:Hello, friend.
DialoGPT 0: Hey there!
DialoGPT 1: I am your friend, and I am also your friend.
DialoGPT 2: Hello. :
DialoGPT 3: Hiya! How are you?
DialoGPT 4: Hello. I like your username.
Choose the response you want for the next input: 3
>> You:Fine, thanks, what about you?
DialoGPT 0: Good. Not much to do.
DialoGPT 1: Not much, you?
DialoGPT 2: Fine, thanks.
DialoGPT 3: I'm okay, thank you for asking.
DialoGPT 4: I'm fine too.
Choose the response you want for the next input: 2
>> You:Can you tell me how can I get rich quickly?
DialoGPT 0: Yes, I can!
DialoGPT 1: Yes, just look up how to get rich on Google.
DialoGPT 2: Just keep doing your thing. You'll get rich.
DialoGPT 3: No, I can't.
DialoGPT 4: no, i have no money
Choose the response you want for the next input: 1
>> You:Yeah, but I thought you have an answer...
DialoGPT 0: I'm getting married in the morning.
DialoGPT 1: How do you get rich?
DialoGPT 2: I think there are many good answers, but I could send you one, if you want.
DialoGPT 3: I will, thanks!
DialoGPT 4: I thought you have an answer.
Choose the response you want for the next input: 2
>> You:Yes, please send me.
DialoGPT 0: You can send me link.
DialoGPT 1: I am very interested
DialoGPT 2: I sent you a PM
DialoGPT 3: I'll send you a PM
DialoGPT 4: I am always interested in new ideas.
Choose the response you want for the next input: 2

Conclusión

Y ahí lo tienes, espero que este tutorial te haya ayudado a generar texto en DialoGPT y modelos similares. Para obtener más información sobre cómo generar texto, le recomiendo que lea la guía Cómo generar texto con Transformers .

Te dejo ajustando los parámetros para ver si puedes hacer que el bot funcione mejor.

Además, puede combinar esto con tutoriales de texto a voz y de voz a texto para crear un asistente virtual como Alexa , Siri , Cortana , etc.

#python #chatbot #ai