« first day (1419 days earlier)      last day (2263 days later) » 
00:00 - 20:0020:00 - 00:00

12:18 AM
~ Wood Wizard
 
 
8 hours later…
8:02 AM
checks for life signs
 
9:01 AM
pokes @sha with @InsaneCat
 
yooooooooooooo
 
Eyyyyyyy :D
What's up?
 
9:18 AM
@Moosebot kill whoever poked @Sha with @Insane
 
user280395
> whoever poked @Sha with @Insane was killed by Col. Mustard with the Wrench in the Billiard Room.
 
@berserk It's good, But not for me, I am not used to this low temperature. Toronto is chill af.
 
Is it above 0 degrees?
 
6 degrees C, Yes!
 
Then it ain't cold! :D
 
9:19 AM
Yeah only for you
 
This temperature is nothing but death God himself
asking us like, "Are you ready?"
 
@PrincessLuna it is, for someone used to 30+
 
@ShadowWizard I am not used to 30+
 
I arrived Canada for honeymoon during their summer, it was 8 during nights and freezing for me. (and wife)
 
9:20 AM
In my last city, temperature is moderate
14-28
throughout the year
 
@TheLittleNaruto oh. Aren't you from India?
 
@ShadowWizard No
 
@TheLittleNaruto avail (?)
 
@TheLittleNaruto that's reasonable, quite like here.
@TheLittleNaruto hehe, since most who arrived from android rooms are, assumed you too. lol :D
 
@ShadowWizard Yes It's reasonable, anything beyond that range is nothing but an invitation of Death God.
 
9:22 AM
So where are you from originally? @TheLittleNaruto
 
@ShadowWizard hehe
 
@TheLittleNaruto lol (?)
 
@ShadowWizard India
 
@TheLittleNaruto pakistan (?)
 
@TheLittleNaruto depends
 
9:22 AM
Wow, it's actually warm outside for once :D
 
@ShadowWizard spacing (?)
 
@TheLittleNaruto umm... but you said "No"... confused
 
@ShadowWizard ..wait! Did I ?
May be because I am high atm.
 
2 mins ago, by TheLittleNaruto
@ShadowWizard No
@PrincessLuna ouch
 
9:23 AM
@ShadowWizard hop (?)
 
If it reach -100, the site will break? @Pri
 
@ShadowWizard why?
 
@PrincessLuna Dude leave that place right now, I repeat leave that place right now.
 
@ShadowWizard no, but the humans who make it will xD
 
@PrincessLuna lol.. why hide the address??
 
9:24 AM
-9 is a nice temp ^^
@ShadowWizard shrugs
 
How is Electricity work there in such temperature ?
 
sending cold wave on @TheLittleNaruto
 
Do you have often blackout ?
 
@TheLittleNaruto no
 
@PrincessLuna That word is already said in the latest 10 words. Please use another. (In case I'm mistaken, run >>rmword no and then >>reply 6727895)
 
9:25 AM
@ShadowWizard runs as fast as I could
 
@TheLittleNaruto here (Israel cold regions, not my own region) water freeze in pipes when it's below 0, but can't see reason why electricity won't work?
 
Electric cars have some problems, some cars don't start at all, but power itself doesn't disappear unless there's a storm and a power line is cut
 
@ShadowWizard Ummm.. may be because of raining ?
 
@TheLittleNaruto well that can happen with 40 degrees too.... ;)
 
Old houses have problems with water freezing in the pipes xD
 
9:26 AM
@ShadowWizard Oh
 
@TheLittleNaruto grumble (?)
 
@TheLittleNaruto um, no
Power lines are protected to avoid short circuiting in rain
Globally
 
That's good
 
Very good
 
@Shadow any good news you have ?
 
9:31 AM
@TheLittleNaruto nope (though no news is good news ;))
 
@ShadowWizard sure (?)
 
@ShadowWizard :)
 
9:54 AM
coughs
 
sneezes
 
gets bored
I was forced to stop the word2vec calculations (I needed sleep :D) so I had to start over
 
Bring berserk
He used to be a fun around
 
@PrincessLuna why don't you leave the computer working while you sleep?
@TheLittleNaruto @ber is always fun
See how:
throwing boobs on @ber
waiting for fun to begin
:D
 
Heat gen is really high when it starts the NN after the word2vec calculations. I have to watch it so it doesn't overheat and supply extra cooling if needed
 
10:00 AM
...waiting
 
@PrincessLuna oh... that sucks. Can't you buy better hardware?
@TheLittleNaruto patience, takes time until @ber wakes up... ;)
 
And until the NN kicks in, the computations are done on the CPU, and I don't want to slow it down by adding more load
@ShadowWizard i could, but it's already good. My computer overclocks the CPU while it's charging (+ 1 GHz) so combined with putting the GPU under the load of a NN creates a ton of heat
The W2V calculations just iterate as fast as the CPU allows, it doesn't generate a lot of heat now
def createTrainingMatrices(dictionary, corpus):
    uniqueWords = list(dictionary)
    allWords = corpus.split()
    totalWords = len(allWords)
    xTrain = []
    yTrain = []

    for i in range(totalWords):
        if i % 100000 == 0:
            print("Finished %d/%d of the words" % (i, totalWords))
        wordsAfter = allWords[i + 1 : i + windowSize + 1]
        wordsBefore = allWords[max(0, i - windowSize) : i]
        wordsAdded = wordsAfter + wordsBefore
        for word in wordsAdded:
            xTrain.append(uniqueWords.index(allWords[i]))
 
@PrincessLuna What does this method do ?
 
It's the main word2vec calculation, the one that currently iterates 3.21 million times xD It creates training data for the embedding matrix
 
@PrincessLuna 3.21 millions ? :O
What is word2vec calculation ?
 
10:11 AM
Yeah. There are a lot of words in the corell movie dataset :D
@TheLittleNaruto converting words to vectors (numbers)
 
@TheLittleNaruto looks like @PrincessLuna try to make something like tensorflow.org :)
 
I'm using TensorFlow
 
oh lol
 
Creating TF would take years xD
 
So what's the final goal? Thought you started from scratch, just for fun.
 
10:13 AM
@PrincessLuna words to vectors ? Care to explain a bit ?
 
@ShadowWizard TF is a ML lib, I need it for the chatbot I want to make
 
@ShadowWizard I know he using TensorFlow
 
@TheLittleNaruto hmm? Who? @ber?
@ber has no idea what's TensorFlow, pretty sure. ;)
@ber has other flows..... :D
 
@ShadowWizard Did I say that? ^^
 
10:14 AM
overflowing @ber with pings
 
@TheLittleNaruto the neural net needs number input, so I'm converting a dataset into vectors which are stored as a vocab which is cross-referenced when getting output and input data for the net
 
@TheLittleNaruto you said "he"... there's no "he" around except me ATM... and @ber and @Dro and some more.... ;)
 
@PrincessLuna sounds like advanced bot! :D
 
@ShadowWizard I meant @PrincessLuna
 
10:16 AM
@TheLittleNaruto lol I know, just teasing :P
 
daymn!
 
@ShadowWizard it is :D I'm still looking for memory modules and how to integrate those, but I think I need NLP for it
 
round kicks @ShadowWizard
 
@PrincessLuna MLP? Ask @Derpy..... ;)
No Little Pony :(
 
10:17 AM
Lol, no. Natural Language Processing
 
@PrincessLuna Example of that cross referencing ?
 
ohhhhh
throwing Natural Language Processing on @TheLittleNaruto, @DroidDev and @ber
 
@PrincessLuna Where is the referencing vectors numbers ?
 
Hold on, lemme go desktop again
 
10:20 AM
Thanks
 
!!/alive
 
lol he's dead and not in room
(kicked out as last hope to revive)
 
Welcome back @rene
 
17
Q: Neural networks - input values

agnieszkaI have a question that may be trivial but it's not described anywhere i've looked. I'm studying neural networks and everywhere i look there's some theory and some trivial example with some 0s and 1s as an input. I'm wondering: do i have to put only one value as an input value for one neuron, or c...

 
I am hoping he was not expecting a welcome...
 
10:23 AM
17
Q: How to convert numpy arrays to standard TensorFlow format?

Keshav ChoudharyI have two numpy arrays one that contains captcha images and another that contains the corresponding labels(in one-hot vector format) I want to load these into TensorFlow so I can classify them using a neural network. How can this be done ? What shape do the numpy arrays need to have? Add...

That should be enough reference
 
Reading...
 
Tensorflow takes number inputs, which are multiplied using node connections, influenced by weights and biases are added to the multiplied result
 
This is too much to digest altogether
 
Tada!!
 
Keras takes numpy input, still numbers
And it's 1/4th done \o/
 
10:26 AM
!!time
 
~ Wed Feb 28 2018 11:26:25 GMT+0100 (CET)
 
I think one of the moderation bots has a neural net
 
!!meow
 
~ WUT?
 
!!cat
 
10:28 AM
 
yo
FATCAT
 
have fun all, have to go now
 
hm, smokedetector doesn't run a NN
 
!!pee
 
~ That is gross ....
 
10:34 AM
same with Belisaurus, Fox9000 (the edit-detector part of it)
 
11:19 AM
coughs
 
11:43 AM
!!status
 
BOT running since: Wed Feb 28 2018 11:24:44 GMT+0100 (CET) for an hour
username         (#msg)
Princess Luna  , seen 23 minutes ago (7)
KennyBOT       , seen an hour ago (6)
rene           , seen an hour ago (5)
InsaneCat      , seen an hour ago (4)
Shadow Wizard  , seen less than a minute ago (2)
TheLittleNaruto, seen an hour ago (1)
 
!!feed cure for @PrincessLuna's cough
 
~ No Pizza?
 
!!feed pizza with little mold
 
~ Come on! Make me something better than pizza with little mold
 
11:44 AM
lol
 
oh
dammit
!!status
I had a massive lead there :D
 
!!cat fat
 
@ShadowWizard I only know these categories: hats, space, funny, sunglasses, boxes, caturday, ties, dream, kittens, sinks, clothes
 
Like 700 events or something
oh well
 
!!cat space
 
11:46 AM
 
I can get it back, I don't have a life after all :D
 
Events are universal, right? so I can do stuff like this
 
You can just leave and join the room too
 
anyways :D
I know
But where's the fun in that? :D
@ProgramFOX did you know there's no fun in python? But there's plenty of it in Kotlin ;P And I think Python definitively agrees :P
 
12:04 PM
XD
 
Wow, who would have thought my jokes were fun? I just thought all my good ones were lost in the void. Yeah, definitively. :>
 
but now you're going into partially repeating yourself
 
Are you sure you're being objective? Because this doesn't look like it
 
Your comment has no value.
 
your argument is null and void
But I'm sure you can get it to function if you want in on the fun
 
12:10 PM
stackalloc
 
You don't have class!
But I'm sure you can interface with something to fix this
lol, when I plug in my computers charger there's a 10 degree spike upward xD
 
At least Android's isUserAMonkey doesn't return true for me - I can't say the same about you!
 
What about isUserAGoat?
@ProgramFOX It returns false for me too. Though I'm getting mixed signals on getSpecies() :P
 
@PrincessLuna hahaha I didn't know about that one
 
there's also Log.wtf (What a Terrible Failure) :D
 
12:17 PM
just like that name itself
terrible
:P
 
lol
The ChatExchange docs are horrible
 
I haven't looked at those in a long time
Well, "horrible", I just checked them, and they're only like 5% complete
 
yeah, exactly. Horrible
 
The 'examples' directory will be more interesting
 
Oh, look, I'm half way in the W2V parsing :D
yeah, I'll look at it
 
12:38 PM
and it's almost up :D running some tests in a sec
 
12:48 PM
dammit
ValueError: invalid literal for int() with base 10: 'users'
 
1:09 PM
@ProgramFOX creative ideas?
 
@PrincessLuna Just restart your program. It's a weird thing that randomly happens when trying to authenticate
 
I have
5 times
different consoles
CMD and Bash
using the wrong password throws a different exception
 
Hmm
what's your code?
 
Ignore the unused imports btw, they're gonna be used later
import numpy as np
import pickle
import tensorflow as tf
import model

import getpass, re, sys, time

from chatexchange.client import Client
from chatexchange.browser import LoginError
from chatexchange.events import MessagePosted, MessageEdited, MessageStarred
from chatexchange.messages import Message
import Config

def start():
    host = "meta.stackexchange.com"
    room = 721

    email = Config.email

    password = getpass.getpass("Password: ")

    client = Client(host)
    client.login(email, password)
room has also had the value '721' as shown in the example
 
that looks okay at a first glance
which line throws the exception?
 
1:19 PM
oh, hold on
nope
thought the naming of the room var caused it, I was wrong
room.join
Traceback (most recent call last):
  File "stackexchange.py", line 31, in <module>
    start()
  File "stackexchange.py", line 25, in start
    room.join()
  File "C:\Users\LunarWatcher\AppData\Local\Programs\Python\Python36\lib\site-packages\chatexchange\rooms.py", line 51, in join
    return self._client._join_room(self.id)
  File "C:\Users\LunarWatcher\AppData\Local\Programs\Python\Python36\lib\site-packages\chatexchange\client.py", line 279, in _join_room
    self._br.join_room(room_id)
  File "C:\Users\LunarWatcher\AppData\Local\Programs\Python\Python36\lib\site-packages\chatexchange\b
 
There's a similar issue using get_me, but that's related to the rep
 
user359686
manual login works
 
try making the room ID a string
 
It was
tried again, same crash
 
1:23 PM
> room = 721
hm
I have no clue then; just suggested that because the examples had it as a string too
Try asking in Charcoal HQ instead, they probably know more about it
 
> room = '721' was the first try
 
aha
 
I used pip to install though
 
Oh
well
I guess you're on an old version then
better use it as a git submodule
 
Wanna guess what I'm gonna say next? :D
 
1:29 PM
"I'm not using git"
then shame on you
 
No
How do I make a git submodule?
 
oh that's better :P
git submodule add ChatExchange https://github.com/Manishearth/ChatExchange.git
git submodule update --init
Then the imports will probably look like from ChatExchange.chatexchange.client import Client, so a bit longer
 
repo URL: 'ChatExchange' must be absolute or begin with ./|../
oh
 
I totally typed that without testing
oh of course
drop the 'ChatExchange' argument
My memory was a bit broken there :P
 
I just put it last and it worked
 
1:34 PM
ah
well in this case it wouldn't have been necessary because it would have created the directory as 'ChatExchange' anyway (I guess that's what the last argument is for then)
 
Inited
@ProgramFOX probably
 
@PrincessLuna celtic (?)
 
Good news: It didn't crash
bad news : Message not sent :/
 
hm
it's asynchronous
so I guess you log out and exit before it's even sent
 
I'm gonna uninstall the pip-installed ChatExchange module
yeah, probably
Yep, it logs out too fast
 
user359686
1:49 PM
Hello, @ProgramFOX
 
user359686
Console test
 
user359686
Hihi, it works \o/
 
user359686
with some manual overriding
 
yay!
 
//hai
right, but I'm stupid and got it to crash so the onMessage callback is broken
1.2 million words left btw xD
 
user359686
1:57 PM
~!end
 
user359686
Hello, @ProgramFOX
 
//hai
 
user359686
@PrincessLuna Hiya!
 
\o/
Anyways, thanks @Pro ^^
 
yw!
 
2:02 PM
now I just have to make it support all the SE sites and Discord
There are API's for everything so it shouldn't be hard
I can get the base bot running without AI support for the time being while I wait for W2V and Seq2seq after that
there's one client per site, right?
 
well, "all SE sites" isn't hard, because there are only three chat domains:
chat.meta.stackexchange.com, chat.stackoverflow.com and chat.stackexchange.com
and yes, there's one Client per domain
 
@ProgramFOX that still means 3 clients though
@ProgramFOX cool, thanks ^^
 
yeah but just making sure you didn't think you'd need 100 clients ;P
 
I know :D
I have another chatbot written in Java already
 
ah cool!
 
2:23 PM
hey @Shadow the Cake Wizard :>
 
2:35 PM
Python has type definitions? Cool!
That saves a lot of time
 
3:19 PM
Oh! That's it!
 
user359686
3:33 PM
Test
 
//help
right
 
user359686
alive
 
//help
 
user359686
@PrincessLuna There's just one command: this one
 
Look at that, it works :D
 
3:47 PM
//unk
//alive
//help
//alive
//alive
//help
 
user359686
@PrincessLuna Sorry, that's not a command I know
 
user359686
@PrincessLuna There's just one command: this one
 
4:58 PM
~ A Shadow in the Streets
 
5:31 PM
dem mass pings
 
Hey @ber :D
 
heya @Lunar
sup
 
Still waiting for the W2V calculation to finish. Had to start over :/ So just doing some maths (got a test next week). How 'bout you?
 
umm nothing like your stuff, just eating icecream :p
 
If I want ice, I'll just go outside the door ;p
 
5:36 PM
lol
but is it creamy? ;]
 
Depends... Do you have the supercomputer donation ready? :P
 
Yes, the consistency of snow is conditional ;p
 
6:08 PM
Depending on the existence of my supercomputer @berserk :p
 
lol
 
6:35 PM
@ProgramFOX btw, the NN is about to kick in
it has < 100k words left
 
Cool!
So I can see it in action in just a couple of minutes?
 
Unfurtunately, no
The Word2vec is just the initial part
This vectorizes the words, creates the vocab and the embedding matrix
after this trains, I have to train the net itself, which is in the seq2seq script
this is just data prep basically
and I still need to hook everything up before it's testable from chat, but that shouldn't take long
 
6:54 PM
here we go, last 10k
 
nice
 
7:09 PM
Dammit, it crashed
Nothing was saved
I'll optimize the method and see if I can speed it up
 
oh no
 
I just wasted 9 hours :/
I'm at an average of 99 iterations per second assuming it took exactly 9 hours
 
Can't you try it with a much smaller dataset, make sure it all works, then use the full dataset?
 
I could, but it's hard finding a small enough one
I think I found a solution to the speed issue
 
@PrincessLuna bug?
@berserk iScream
 
7:15 PM
No associated word found for iScream.
 
Typo as a result of brainlag from the multitasking
 
lol
 
>>continue iScream ice_cream
 
@ShadowWizard ice cream (?)
 
@FOX9000 rene
 
7:16 PM
@ShadowWizard SEDE (?)
 
@PrincessLuna Can't you split your current dataset in pieces?
 
I could, but file splitting is a pain
But I can remove some redundant vars in the loop though
 
It sure is, but wasting 9 hours is more of a pain :P
 
I need to get it back up to a high speed iteration at ~10k per second (or at least 5k)
With little memory allocation that should be possible
I need to do some dry testing
 
5K/sec sounds much better than 99/second yeah
 
7:25 PM
Basic appending iterates 2 million times in 0.26 seconds
with multiplication .37
exponents significantly reduces speed
10k in 5 seconds, 2k per second
I can prevent word duplication
I can do some multithreading
But that would be a pain to reassemble
well, this explains it
I appended all the previous words back to the list
but those are needed to get actual sentences, I think
ok, cut the var and loop the new ones excluding the existing ones, which should speed it up
10k in 6 seconds
but it slows down
ok, looks like multithreading is the only way to speed it up more
 
7:59 PM
It did speed it up at least
 
00:00 - 20:0020:00 - 00:00

« first day (1419 days earlier)      last day (2263 days later) »