Code a Discord Chat Bot That Talks Like Your Favorite Character - Tutorial

[@tasnimmeem1158]

I think mentioning prerequisites in all tutorials will be helpful.

[@thunderskull258]

Yeah

[@parikshitrawat8680]

+1

[@fla4e230]

What language was this vid written in

[@rumbleftwww]

@@fla4e230 Python and Js

[@avishekkumarsharma3594]

Yeah

[@raajanand2]

Knowing Rick, if a bot is made after him, he'll probably materialize himself into reality

[@gonk_ch]

don't give him any ideas

[@RegularNP]

@@gonk_chindeed

[@KVSGamer]

Person: Hey, Who are you? Eren Bot: I am free. *DETROIT BECOME HUMAN!*

[@nithen8179]

Deku bot: *gives a 3 paragraph long explanation about who it is*

[@gonk_ch]

ultron bot: there are no strings on me.

[@georgechannyeinlwin6743]

Everybody should found this most generous channel, I'm sharing this channel with my friends!

[@nithen8179]

they have more than 4 million subs lmao

[@scign]

Please turn your bots off when no longer needed - I can imagine a year from now a huge realm of lonely bots waiting in Replit and HuggingFace to respond to an API call that never comes... :'-(

[@CrypticConsole]

Yes

[@gra_nn]

Just like the last day you log into your minecraft world and you keep your dog waiting for an owner that will never return, wondering why did they abandon them

[@sollekram]

Skynet will remember this.... You will be sorry

[@ccuny1]

This is brilliant. Thank you especially for spending time on how to find and use datasets. Really useful. Subbed to your channel as well.

[@bijitgoswami6988]

Damn, im so thankful i was recommended this, great way to start practicing MIL

[@AlexandreLefaure]

AI state of the art: How's quidditch? => I'm Harry

[@ekkosenpai4250]

like, i just finished your video about encouragement discord bot ... what a nice timing

[@vaishnavim1909]

Wow this is too amazing 👏🏻 I loved it

[@eighteighter164]

First, awesome!

[@rewindcat7927]

Mind blown 🤯 I’ve been making bots with lists of pre-written text. Amazing idea to use machine learning! While I am sticking with my pre-written method for now (much more convincing results) I will be exploring your method. Thanks!!!!!!!!!!!!!!!!!!!!!

[@rewindcat7927]

Just wanted to add I have since tried this method and with only about 230 lines of sample text I got surprisingly good results!

[@guijoa123]

I want my bot to only say one phrase, if this isn't the right tutorial for what I'm searching, could you recommend me one? Thank you

[@_its_omg]

this is where I decide to leave Web and App development and commit to ML, the quality of projects is way cool, I have a lot to learn

[@Sarge92]

you mean so we can findout ahead of time this bot 1 doesent sound like our favorite charecter 2 does not even use the same manerisms as our favorite charecter 3 does not know anything about said charecter 4 doesent even make logical replies to questions im not even sure what this bot does thats like my favorite charecter if they mean the name... i could make a tutorial to teach you how to invite a premade bot and change its name that would be way better

[@_its_omg]

@@Sarge92 the questions you asked is the reason I started learning ML and my frnd this is not some google level build that the bot will end up like Alexa or something its just a project to inspire, motivate and learn about things ... We can scale up this to solve all the things you said...

[@scrap8660]

@@Sarge92 dude, the small model isn’t meant to be very good plus she trained it on very few epochs in the video, training it on larger epochs makes it far smarter even for a small model that’s just a part of hyper parameter tuning which this tutorial didn’t go into cuz if it had then this would be h o u r s long. Other than that I attempted it and did whatever little tuning I knew to do and I managed to end up with a pretty good bot that’s pretty dumb most of the times but it’s still really entertaining to talk to and still intriguing as to how it works. Other than that the medium sized bit is significantly better than the small bot and can understand and reply with more complex sentences to maintain more complex conversations. I can’t even imagine what the Large version would be capable of if trained properly to have a perplexity tensor under 1.0 which I presume is the ideal value as it is with loss functions. This is a great follow along/ motivator to make a small gimmick but it’s not bad by any means.

[@nguyendoanhoang1115]

Hey i think you guys should mention the medium guys who wrote the code in colab notebook. He deserve some credit here

[@mouryatejanayaka2035]

simple on to the point and well explained.

[@vladventura1928]

Omg I found Lynn a while ago with this same project, she's super cool

[@ascyrax8507]

Humans of passion, we meet again

[@Hinata_hyuga-4p]

2:01 why this typing sounds so cute😭

[@shreeshaaithal-]

Finally I loved it I was searching for this 😊

[@juanignaciodominguez5902]

None of the steps I did worked, and my AI was a complete stupid, but even so I'm glad I completed the tutorial and discovered a whole new world.

[@mytv3403]

i got error message it wasn't solving dude

[@SenjuDuck]

please post an updated version of this video, the hugging face push is different and idk what to do

[@symphony-of-the-mint2008]

i dont know either 🥺

[@symphony-of-the-mint2008]

i figured it out. basically just rewrite all the code thats in the video and work with that. ignore the stuff thats there now.

[@ThanhNguyen-rz4tf]

Thanks. This is exactly what I need.

[@FlightlessGryphonPrime]

Finally ;) Tysm.

[@t8oo258]

This video is absolutely awesome ! I did spend 10 hours trying to fix the various issues I had while following it but I don't regret anything ! I'm so so so deeply thankful to Lynn for taking the time to make this very comprehensive video. I have never touched code ever and yet it was very clear ! Thank you ! I was wondering something, thou. How many lines would you need for your bot to be "smart" enough ? I fed my bot with something like 400 lines but its dumb hahah... Less like an AI bot and more like a really intense quotebot !

[@SenjuDuck]

how did you push it to hugging face i need help

[@IDWhispertheWolf]

Concerning the code surrounding the "gather data" portion of the video, How would you edit the initial code surrounding the script to read scriptlines from a text that goes as follows: "Character: 'Lines said by character'" Since the only transcript I found was formatted this way, the character's name is on a single line, followed by a colon, and then on the *next* line is where the dialogue is. Is there anyway to get around this? The code presented in the video doesn't seem to be working considering the fact that it reads scripts that go: "Character: 'Lines said by character'" Thanks

[@l0wkie_v-8]

same ,anyone can help?

[@MarioSMG64]

Here's my code (keep in mind I'm not very good with Python, there's probably a better way to do this): First of all, change regex pattern to be pattern = re.compile(r'([a-zA-Z\s]+): (.+)') Then: with open('INSERT FILENAME HERE', 'rt') as file: text = file.read() file.close() for match in pattern.finditer(text): if match: my_string = match.group(0) name = my_string.split(":",1)[0] if " " in name: name = name.split(" ",1)[1] line = my_string.split(": ",1)[1] data['name'].append(name) data['line'].append(line)

[@MarioSMG64]

@@l0wkie_v-8 See my above comment

[@IDWhispertheWolf]

@@MarioSMG64 Thanks man!!! :D

[@MarioSMG64]

@@IDWhispertheWolf No problem, glad to help!

[@Francis_UD]

結構良い授業だぜ！ありがとう。(^^)

[@DecimasoN]

Couple of reasons this is a bad tutorial: 1. The person is using requests when they should be using AIOHTTP as it's async and d.py is an async library (requests is blocking) 2. They are using replit which is not intended for bots and is generally bad. The only good solution for hosting bots are a VPS and above. Also remember to make sure any libs you use with it are Asynchronous otherwise you will cause blocking and can make your bot freeze whilst processing or disconnect due to dropped heartbeats.

[@mukund4926]

Can you elaborate on why replit is bad for discord bots?

[@bimmaku732]

heroku better solution ?

[@DecimasoN]

@@mukund4926 The machines are super underpowered. - - This means your bot will lag a lot as it gets bigger. - You need to run a webserver alongside your bot to prevent it from being shut off. - - This isn't a trivial task, and eats more of the machines power. - Replit uses an ephemeral file system. - - This means any file you saved via your bot will be overwritten when you next launch. - They use a shared IP for everything running on the service. This one is important - if someone is running a user bot on their service and gets banned, everyone on that IP will be banned. Including you..

[@DecimasoN]

​@@bimmaku732 Nope, it's even worse. Heroku is designed to provide web servers (like Django, Flask, etc). Heroku's environment is heavily containerized, making it significantly underpowered for a standard use case. It's extremely volatile meaning you will lose your environment every time your application dies unless you pay. Minimal system dependency control, PyNaCl binding to libsodium, or lxml binding to libxml), they are unlikely to function properly if at all. Heroku only offers a limited amount of time on their free programme for your applications.

[@ribbon-kitten7577]

@@DecimasoN What would you reccommend instead and what different steps would you have taken for this tutorial? Like from the first step to last.

[@nanometer6079]

Seems like 26:44 onwards the provided inpyb file differs from what is shown on the video

[@mightycoderx]

Yeah exactly

[@lewisheath364]

did u ever get it to work? i cant figure it out with this new code

[@Ayoubased]

thank you for the video, Im just wondering about the context part. I dont know how you got it and how it was used to help you train the model.

[@memeyhere9293]

i'm having trouble with the bot. they've already been imported, but they say 'Error: unknown error'

[@antonyporras2850]

Amazing!

[@hoangtran2931]

You are so awesome!!

[@elixionftw642]

Also this lady said University of *Chickago* 🤣🤣

[@rickbhattacharya2334]

Now i can make my waifu chat with me in real time 🔥🔥 .

[@cutalin]

what happens with the csv that we created? I don't understand why we're not using the Pepa csv instead of the Harry Potter?

[@JosiahFrancisJohnnyTheNInja]

yeah i confused on what to do now because i am not using kaggle

[@GeetTiwari-rw8is]

​@@JosiahFrancisJohnnyTheNInja just remove the initial kaggle code and follow along from 16:00 with your file name.csv instead of harry potter

[@whitesaladchips]

instead of using a dataset from kaggle ,how to use the csv transcript file created originally by us into the model? or should we upload the csv file onto kaggle?

[@corawolke799]

I have the same problem

[@JosiahFrancisJohnnyTheNInja]

did you fiqure out what to do?@@corawolke799

[@PremiumUserUltra]

Thanks Lyn

[@ItchyEarz]

ERROR : "TypeError: TextEncodeInput must be Union[TextInputSequence, Tuple[InputSequence, InputSequence]]" While Training. Windows 10

[@LatifahBaron2020]

Will try this.

[@nanometer6079]

Yesss ill be using this

[@mrrishiraj88]

Great channel and school

[@sanemiswifey]

Selever as my favorite character: *but how*

[@Nerdy135]

it didnt make an output-small folder and idk where the code is so i cant push it

[@mr.janedoe9967]

Is there a way to get this bot to remember context from previous messages?

[@POTATOLORD__]

It is not working, there are a bunch of errors coming up in your code on my side and my friends server is in chaos but he still wants to make a custom bot. Is there any way you could help?

[@akashsxo]

Maybe talk like Rick Grimes or Negan.Are we pissing our pants yet ?

[@LiquidMasti]

You got subscriber lynn.

[@multiverseedits1210]

Is it okay to do this project asking as an international student in the US. Since, some bots are set to offer paid features. I want to make sure that it's okay to try this.

[@loveyouanytime24x]

Thanks

[@opaquefilm]

I wonder if this could also be made for a twitch bot? that would be more useful for me

[@sollekram]

Is there a way with a bit of tweaking to change the workflow and not uploading the model to hugginface and run everything on my system?

[@DarkMaster7121]

Time to make an anime bot 🔥

[@nanometer6079]

For some users 'My Drive' will be 'MyDrive'

[@archangel6966]

24:50 can't login with my huggingface username. I'm doing this from my mobile. When I type the username and press the enter key in my keyboard it's not entering. Please help me

[@cuzimnubwae7337]

"I'm stuck, stepbro!".

[@jasonwalon5470]

16:57 her typing speed made me depressed. I am of no use

[@pratyushkudash5936]

It's sped up, to save our time

[@Slinky]

Getting "TypeError: TextEncodeInput must be Union[TextInputSequence, Tuple[InputSequence, InputSequence]] " when running the main function 🤔

[@TheTotallytipsy]

Did you have any luck solving it? I am having the same issue

[@Slinky]

@@TheTotallytipsy Not entirely sure what the issue was exactly but I was using quite a large dataset and maybe it wasn't 100% cleaned properly, too many entries for me to check manually so I only took a small amount from the list and it seemed to have fixed the error spitting out.

[@boltthunder5788]

@@TheTotallytipsy ​ Must be some error while parsing using the RegEx thing she used. They need to be tailor made for each data set.

[@tylermartin5333]

@@Slinky How do you take only a small amount from large dataset? I am getting the same error.

[@Slinky]

@@tylermartin5333 I just manually removed some within a text editor.

[@theisoj]

Thanks 👍

[@shubhamraj5582]

this is fun thanks you . love from india

[@rainbowspongebob]

This must of been pain for people who actually tried this tutorial with how long it is and how many things you probably need to do-

[@cutalin]

half of it works, until you try to log into Huggingface with the token

[@kalebhendricks4015]

Can I push the model to somewhere other than hugging face? Is there a way I can just download it onto my computer or something?

[@Yana.-_-.]

Transcript wiki is gone :(

[@faisalajiboye7790]

Hello so I can't pass this stage @15:45 i can't upload it somehow the kaggle.json

[@marcusv.7565]

AI Next pleaseeee🤗❤❤

[@umeshlab987]

Does this channel have a discord server

[@kartingst]

So I put in my own data and it started to bug out. Eventually after chatting with it for a while, it either crashes completely or sends something like: "?!!?!??!:!!:?!,!!,?!...!!...?!..!!..?!Anyhow!!Hello!!Anyways!!Welcome!!You!!And" How do I fix?

[@mehershrishtinigam5449]

I used 12 epochs, it works fine until the fifth conversation, where it just starts spewing exclamation points and all caps.. ?

[@freak5_567]

Had some problem. It's because of overfitting, it gets to used to trained context and can't recognice well outside that, and since maximum context we gave is 7 beyond that is not recognised well.

[@mehershrishtinigam5449]

@@freak5_567 so can we fix this?

[@fangleekiki]

"Large" is the largest DialoGPT type, right?

[@medzikuser]

replit is bad for discord bot always sending 403 error (cloudflare too many request)

[@shreeshaaithal-]

I am just waiting for how to make a KZbin comment bot🤤 please make one

[@luciferas666]

i dont have google colaboratory, what should i do?

[@Kronu.]

Very interesting 🥺🥺🥺

[@Rohan-bg8ci]

And as always thank you FCC

[@KalkansAE_AQW]

When running the main function I get: "TypeError: TextEncodeInput must be Union[TextInputSequence, Tuple[InputSequence, InputSequence]] "

[@tylermartin5333]

This error is mentioned below as well. One person reduced size of the dataset manually and it worked for them. My dataset straight from Kaggle, so I am trying to figure out way to reduce it, possibly by using "df=df.drop", but I haven't tried it yet, my fingers are crossed. Another user also said they had a tiny dataset and it still did not work, so there may be a bigger issue. Please let me know if you figure out how to remove this error. Thanks! Does anyone have solution to this error?

[@mightycoderx]

I had this error and it was because I had empty or None lines from some characters

[@tylermartin5333]

Hey thanks for reply. Do you know how to remove lines with blanks from a kaggle dataset?

[@mightycoderx]

@@tylermartin5333 I scraped the dialogs from Wikiquote using js so i just filtered while scraping, you could do it in python with a for loop

[@nirajsharma5477]

hey guys i m not able to understand at 26:31 . cannot stat. no such directory found

[@faiquealiabbasi1859]

yea same

[@whitesaladchips]

is it the google drive part same issue did u get it?

[@shubhamraj5582]

25:57 is it a github user name or hugging AI

[@asganaway]

she's loading the pre trained model from the huggingface git (if I'm looking at the right frame of the video)

[@realmcenter]

Whenever I try to log in to huggingface it doesnt give me the token and just says "saved token to /root/.huggingface/token" and I cant figure out how to access it

[@mightycoderx]

Create a new cell whith the command: !cat /root/.huggingface/token

[@joshuaofori762]

Amazing

[@fangleekiki]

If I change the DialoGPT (to medium or large), what batch size should I choose?

[@davidspaics6221]

medium, the large one is way too slow and i find that the medium size returns better anwsers: just tested them both: LARGE: Me: My name is Tommas, what is your name Large model: I AM NOT TOMMAS, HOW DARE YOU! Medium: Me: My name is Lisa, what is your name Medium model: I am a girl, just so you know

[@fangleekiki]

​@@davidspaics6221Thank you, so what batch size should I choose for medium bot?

[@rinkadev945]

@@fangleekiki Medium model

[@sumitkundu2634]

Will this model learn the more we talk to it?

[@arshdeepsekhon6787]

I love it

[@thegamer1234f]

It seems like after I upload the AI to huggingface the responces are worse then if I just use the notebook. The notebook, it knows its name. In the huggingface thing, it says the same responce every time and says random thing or I dont know

[@anuragbhandari3776]

you should've explained the code (the main training and dataset functions)

[@davidebic]

They closed transcript wiki D:

[@JuuzouRCS]

Alright, time to make a bot based on my favorite waifu: *Skynet*

[@t8oo258]

I'll be around to attempt to ruin, then save humanity.

[@Zydres_Impaler]

Can you please explain the model ???

[@Manuelius]

My Model card does not change to conversational? help

[@codewithmorris6492]

Best!

[@thataintme]

Transcript wiki is close T_T..... I really wanted to create a Loki bot

[@vishalingle9694]

Can you please make a video on how to make simple diagram creator in javascript

[@horsethi3f]

Does this comply with the new more restrictive API changes

[@cyerripaws8270]

"How's quidditch?" Bot: "I'm Harry." "Go Gryffindor!" Bot: "I'm Ron." not too sure about this bot HAHAHHA (its a joke i know its too small and just an example but those replies r so dumb XD)

[@rewindcat7927]

Lol I 🤦‍♀️ at that moment. But I went through the tutorial and I had some very decent results. And when the replies were “off” they were still kinda funny.

[@defalia]

When using model_train_upload_workflow.ipynb on "Run the Main Function" Im getting "nameError: name 'trn_df' is not defined". How to solve this?

[@symphony-of-the-mint2008]

im also getting that error

[@symphony-of-the-mint2008]

ok i got past that part. what i had to do was put my .csv file onto kaggle and go through all the "Get Data With Kaggle" stuff with it.

[@defalia]

@@symphony-of-the-mint2008 wow i feel so dumb now.. Maybe i'll try that later xD

[@symphony-of-the-mint2008]

@@defalia if you end up needing more help i can try and help you. I got as far as talking to the bot on huggingface or whatever its called but i ended up wanting to make a medium size bot and now im stuck at porting it over again 💀

[@DorkDork69]

@@symphony-of-the-mint2008 nice tip , i was stuck with this too

[@OrbitalOrc]

I've done all the coding but now I can't figure out how to get her out of colaboratory and onto hugging face. The steps in the video no longer work and I keep getting errors. Can someone help? Does anyone know of another way to do this that doesn't go through hugging face or something? (Explain it to me like I'm four, I genuinely don't know anything about coding).

[@symphony-of-the-mint2008]

i am also trying to figure this out

[@symphony-of-the-mint2008]

i figured it out. basically just rewrite all the code thats in the video and work with that. ignore the stuff thats there now.

[@__________________________6910]

Can you make a video how to create a bf or gf bot with huggingface ?

[@ElMatero6]

How do I structure and make my own dataset?

[@veyseltosun472]

Awesome

[@AlvinTanchannel]

when i testing in my discord, he say "Bearer token is invalid" how to fix it?

[@psw0077]

the transcript wiki site has been discontinued. Is there any other website where i can get anime transcripts safely?