AlphaDev Solves Algorithms Faster Than Humans

Deepmind, Google’s AI branch, has just presented AlphaDev, a Reinforcement Learning agent that can discover faster sorting algorithms on its own.

AlphaDev’s advanced computer science algorithms outperform decades of human benchmarks from engineers and scientists.

These new algorithms are already part of two standard C++ coding libraries and are being used trillions of times per day by programmers worldwide. In addition, AlphaDev has also demonstrated potential in enhancing other essential computer science algorithms, such as hashing.

This represents a significant step towards using AI to optimize the world's code, one algorithm at a time.

Source:

https://www.deepmind.com/blog/alphadev-discovers-faster-sorting-algorithms

Repair slave in MySQL GTID replication, after fatal error 1236 (slave has more GTIDs than the master)

For repairing a broken slave after fatal error 1236.

Last_IO_Error: Got fatal error 1236 from master when reading data from binary log: 'Slave has more GTIDs than the master has, using the master's SERVER_UUID. This may indicate that the end of the binary log was truncated or that the last binary log file was lost, e.g., after a power or disk failure when sync_binlog != 1

slave# stop slave for channel 'mymaster';

slave# reset slave for channel 'mymaster';

slave# set gtid_next='automatic';

slave# change master to master_auto_position=1 for channel 'mymaster';

slave# start slave for channel 'mymaster';

slave# show slave status for channel 'mymaster'\G

OpenAI ChatGPT

ChatGPT, which stands for Chat Generative Pre-trained Transformer, is a chatbot developed by OpenAI. ChatGPT is built on top of OpenAI's GPT-3.5 family of large language models, and is fine-tuned with both supervised and reinforcement learning techniques.

ChatGPT was launched as a prototype in November 2022, and quickly garnered attention for its detailed responses and articulate answers across many domains of knowledge.

It was fine-tuned on top of GPT-3.5 using supervised learning as well as reinforcement learning. Both approaches used human trainers to improve the model's performance. In the case of supervised learning, the model was provided with conversations in which the trainers played both sides: the user and the AI assistant. In the reinforcement step, human trainers first ranked responses that the model had created in a previous conversation. These rankings were used to create 'reward models' that the model was further fine-tuned on using several iterations of Proximal Policy Optimization (PPO). Proximal Policy Optimization algorithms present a cost-effective benefit to trust region policy optimization algorithms; they negate many of the computationally expensive operations with faster performance. The models were trained in collaboration with Microsoft on their Azure supercomputing infrastructure.

ChatGPT is part of a series of releases around GPT 3.5 that are highlighting some of the capabilities of the upcoming GPT-4 model. One of the key differences of ChatGPT with previous models is its ability to follow instructions. This is powered another model called InstructGPT which OpenAI quietly unveiled at the beginning of the year.

Large language models like GPT-3 are often used to follow instructions to execute user’s tasks. However, quite often, these models generate toxic or untruthful outputs that are not related to the input instructions. This is mostly due to the fact that models like GPT-3 are trained to predict the next word in a sentence rather than to execute a specific task. This is precisesly the problem OpenAI tried to address with InstructGPT, a language model that builds upon GPT-3 language capabilities but improves it its ability to follow instructions.

More about InstructGPT

Training language models to follow instructions with human feedback.

Making language models bigger does not inherently make them better at following a user's intent. For example, large language models can generate outputs that are untruthful, toxic, or simply not helpful to the user. In other words, these models are not aligned with their users. In this paper, we show an avenue for aligning language models with user intent on a wide range of tasks by fine-tuning with human feedback. Starting with a set of labeler-written prompts and prompts submitted through the OpenAI API, we collect a dataset of labeler demonstrations of the desired model behavior, which we use to fine-tune GPT-3 using supervised learning. We then collect a dataset of rankings of model outputs, which we use to further fine-tune this supervised model using reinforcement learning from human feedback. We call the resulting models InstructGPT. In human evaluations on our prompt distribution, outputs from the 1.3B parameter InstructGPT model are preferred to outputs from the 175B GPT-3, despite having 100x fewer parameters. Moreover, InstructGPT models show improvements in truthfulness and reductions in toxic output generation while having minimal performance regressions on public NLP datasets. Even though InstructGPT still makes simple mistakes, our results show that fine-tuning with human feedback is a promising direction for aligning language models with human intent.

Ref. Cornel University

Pathways Language Model (PaLM): Scaling to 540 Billion Parameters for Breakthrough Performance

Google AI introduced the 540 billion parameter Pathways Language Model. Trained on two Cloud #TPU v4 pods, it achieves state-of-the-art performance on benchmarks and shows exciting capabilities like mathematical reasoning, code writing, and even explaining jokes. https://goo.gle/3j6eMnK

PaLM demonstrates the scaling capability of the Pathways system to thousands of accelerator chips across two TPU v4 Pods by training a 540-billion parameter model efficiently with a well-studied, well-established recipe of a dense decoder-only Transformer model. Pushing the limits of model scale enables breakthrough few-shot performance of PaLM across a variety of natural language processing, reasoning, and code tasks.

PaLM paves the way for even more capable models by combining the scaling capabilities with novel architectural choices and training schemes, and brings us closer to the Pathways vision:

<<Enable a single AI system to generalize across thousands or millions of tasks, to understand different types of data, and to do so with remarkable efficiency.>>

How to Kill MySQL Queries

The command SHOW PROCESSLIST lists all the open connections to the server

 mysql> SHOW PROCESSLIST;

In order to kill a thread completely, use the KILL command followed by the thread ID returned by SHOW PROCESSLIST:

mysql> KILL 1234;

Repair slave in MySQL GTID replication, after fatal error 1236 (master has purged binary logs)

For repairing a broken slave after fatal error 1236.

Last_IO_Error: Got fatal error 1236 from master when reading data from binary log: 'The slave is connecting using CHANGE MASTER TO MASTER_AUTO_POSITION = 1, but the master has purged binary logs containing GTIDs that the slave requires.'

slave# reset master;

slave# source /pathto/dump.sql;

slave# start slave for channel 'mymaster';

slave# show slave status for channel 'mymaster'\G

Here is another useful tutorial, if the problem persists

https://www.percona.com/blog/2013/02/08/how-to-createrestore-a-slave-using-gtid-replication-in-mysql-5-6/

Demo of OpenAI GPT-3

Incredible demos of OpenAI GPT-3, watch this video!

https://www.youtube.com/watch?v=8psgEDhT1MM

Mind-blowing conversations with OpenAI GPT-3

https://medium.com/@kirkouimet/my-mind-blowing-conversations-openais-latest-ai-gpt-3-235ba5fb9453

Folding@Home contro il COVID-19

Folding@Home contro il COVID-19: una potenza mai vista per il calcolo distribuito. Contribuiamo tutti col nostro PC 💪💪

#uniticontroilcoronavirus

https://foldingathome.org

Partecipare al progetto Folding@home è semplicissimo. Basta recarsi sul sito ufficiale, scaricare il client (funziona su Windows, Mac e Linux) e, in automatico, parte della potenza del nostro PC verrà dedicata alla ricerca. L’unione fa la forza.

Repair slave in MySQL GTID replication

Run these commands on MySQL shell, to find "where" replication failed and was broken.

Stop the slave.
SET GTID_NEXT="AUTOMATIC";
stop slave;

Find the point where replication broke.
show slave status\G;

Now find "Executed_Gtid_Set". Example value:
Executed_Gtid_Set: 44d966cc-d84d-11e8-be8d-20677cd3f384:1-2872

Increase of one step to 2873
SET GTID_NEXT="44d966cc-d84d-11e8-be8d-20677cd3f384:2873";

Now start the slave.
start slave;
show slave status\G;


Here is another useful tutorial, if the problem persists
https://www.percona.com/blog/2013/02/08/how-to-createrestore-a-slave-using-gtid-replication-in-mysql-5-6/

Continue the loop until no error occurs. Enjoy

Choosing a visual to represent data

How do you choose a visual to represent data?

Here is a handy diagram that can be used to determine the right chart type to use in your data visualization.

How To Secure MySQL Replication Using SSH

https://www.digitalocean.com/community/tutorials/how-to-secure-mysql-replication-using-ssh-on-a-vps

http://makandracards.com/makandra/573-mysql-replication-how-to

http://www.somacon.com/p567.php


Bi-directional SSH tunnel

local$: ssh -N -R 33061:localhost:3306 remote_host_ip

remote$: ssh -p 33061 localhost


Create/restore a slave using GTID replication
https://www.percona.com/blog/2013/02/08/how-to-createrestore-a-slave-using-gtid-replication-in-mysql-5-6/


MySQL Multi-Source Replication
http://www.xenialab.it/meo/web/white/oracle/myrepl_multi.htm

If we have a chain of slaves, it's important to enable slave update logging, on config file:
log-slave-updates = 1

Unable to install Certbot

Error on installing Certbot for Lets Encrypt

# apt-get install -t jessie-backports certbot python-certbot-apache Reading package lists... Done E: The value 'jessie-backports' is invalid for APT::Default-Release as such a release is not available in the sources

Solution:
You should first enable backports, see https://backports.debian.org/Instructions
http://guide.debianizzati.org/index.php/Il_repository_Backports

Recupero dati da scheda micro SD, e non solo...

Se devi recuperare dei file cancellati da un hard disk, una chiavetta USB o una scheda SD, ti consiglio vivamente di rivolgerti a PhotoRec. Si tratta di un software gratuito e open source che permette di recuperare moltissimi tipi di file – immagini, video, documenti ecc. – in pochi click. È compatibile con tutti i principali sistemi operativi per computer: Windows, macOS e Linux. Viene fornito in coppia con TestDisk, un altro software gratuito che permette di recuperare le partizioni perse e di rendere avviabili i dischi che non lo sono più: è molto efficace, ma al momento non dovrebbe servirti.

Per scaricare PhotoRec, collegati al sito del programma e clicca sulla voce corrispondente al sistema operativo installato sul tuo PC (scegli ”Windows” indipendentemente se usi una versione di Windows a 32 o 64 bit). Al termine del download, apri l’archivio zip che contiene PhotoRec e TestDisk, estraine il contenuto in una cartella di tua scelta e avvia l’eseguibile qphotorec_win.exe.

Nella finestra che si apre, espandi il menu a tendina collocato in alto e scegli l’unità relativa al dispositivo da cui intendi recuperare i dati. Clicca quindi sulla voce corrispondente alla partizione principale del drive e clicca sul pulsante File Formats per selezionare le tipologie di file che vuoi recuperare (meno tipologie di file selezioni, minore sarà il tempo impiegato dal software per svolgere il suo lavoro).

Successivamente, metti il segno di spunta accanto alle opzioni Whole e FAT/NTFS/HFS+/ReiserFS, fai click sul pulsante Browse e seleziona la cartella in cui PhotoRec deve salvare i file che riesce a recuperare. Per concludere, clicca sul pulsante Search e attendi che venga completato il recupero dei file cancellati, i quali verranno salvati automaticamente nella cartella selezionata.

ripristino MySQL corrotto

Per ripristinare una istanza di MySQL corrotta ho dovuto fare nel seguente modo:

1. Se non è disponibile un dump dei dati, cercare di fare avviare il servizio di MySQL
1. Se non si avvia, modificare my.cnf aggiungendo innodb_force_recovery = 6, sotto [mysqld], partendo da un valore basso fino ad un massimo di 6
2. Eseguire il dump di tutto: mysqldump --force --compress --triggers --routines --create-options -uUSERNAME -pPASSWORD --all-databases > /backup/dump.sql
3. Fermare il database: mysqladmin -uUSERNAME -pPASSWORD shutdown
4. Rimuovere tutti i file dalla directory dei dati: rm -rf /path/dati/
5. Ricreare la directory dei dati con permessi 755 ed owner mysql:mysql
2. Inizializzare il database col comando shell#: mysqld --initialize
3. Avviare il servizio senza il controllo credenziali: mysqld --skip-grant-tables
4. Importare i dati col comando shell#: mysql -uroot --compress -p < /backup/dump.sql
5. Modificare la password di root
1. # mysql -uroot mysql
2. $mysql> UPDATE user SET Password=PASSWORD('my_password') WHERE user='root';  oppure se non funziona...
3. UPDATE user SET authentication_string=password('my_password') WHERE user='root';
4. $mysql> FLUSH PRIVILEGES;
6. Riavviare il servizio: service mysql restart

Resize partition in Debian

How to resize partition to full disk size

umount /dev/vdb1
cgdisk /dev/vdb
- remove partiton
- create a new partition
partprobe
resize2fs /dev/vdb1

If cgdisk is not installed
apt-get install gdisk

Viv - incredible next gen AI assistant

Install SSL Certificate on Apache for Debian

Prerequisites
OpenSSL installed and updated (should be installed by default)

apt-get update 

apt-get upgrade openssl

Apache will be our HTTPS server. To install it, run the following:

apt-get install apache2

Enable the Apache SSL module

a2enmod ssl

The default Apache website comes with a useful template for enabling SSL, so we will activate the default website now

a2ensite default-ssl

Configure Apache to Use SSL

Edit server configuration file

nano /etc/apache2/sites-enabled/default-ssl.conf

Locate the section that begins with <VirtualHost _default_:443>.
Add a line with your server name directy below the ServerAdmin email line. This can be your domain name or IP address:

ServerAdmin webmaster@localhost 

ServerName example.com:443

Find the following two lines, and update the paths to match the locations of the certificate and key. If you purchased a certificate or generated your certificate elsewhere, make sure the paths here match the actual locations of your certificate and key:

SSLCertificateFile /etc/apache2/ssl/apache.crt 

SSLCertificateKeyFile /etc/apache2/ssl/apache.key

Restart Apache to put these changes into effect

service apache2 restart

Prestashop friendly url

Problemi con i friendly url di Prestashop? Se il webserver è Apache2 fare come segue...

Abilitare il modulo mod_rewrite:

a2enmod rewrite

Modificare il file 000-default.conf:

nano /etc/apache2/sites-enabled/000-default.conf

Nel blocco  alla riga 1 inserire:

 <Directory /var/www/html>  
         Options Indexes FollowSymLinks MultiViews  
         AllowOverride All  
         Order allow,deny  
         allow from all  
 </Directory>  

Riavviare Apache:

service apache2 restart

Calculating Median with MySQL

Query per calcolare la Mediana in MySQL

select avg(test_val) as median from
(
select @counter:=@counter+1 as row_id, t1.test_val 
from test100 t1, (select @counter:=0) t2
order by t1.test_val
) o1
join
(
select count(*) as total_rows
from test100
) o2
where o1.row_id in (floor((o2.total_rows + 1)/2), floor((o2.total_rows + 2)/2))

Calcolare la Mediana con MySQL

Query per calcolare la Mediana in MySQL

select avg(test_val) as median from
(
select @counter:=@counter+1 as row_id, t1.test_val 
from test100 t1, (select @counter:=0) t2
order by t1.test_val
) o1
join
(
select count(*) as total_rows
from test100
) o2
where o1.row_id in (floor((o2.total_rows + 1)/2), floor((o2.total_rows + 2)/2))