BASED GPT - Building an AI That Says All The Crazy Sh*t You Won't Say at Work

DISCLAMER:....idk have fun with this one, don't do anything I wouldn't do.

💰Support the stream!
CashApp: https://cash.app/$jawerty210
Venmo: https://venmo.com/jawerty210
Buy Me a Coffee: https://www.buymeacoffee.com/bjGHFVW355

Join The Discord: https://discord.gg/dv4TSzsk27

Schemeology Podcast: https://youtube.com/@schemeology

My Socials:
Github: https://github.com/jawerty
Twitter: https://twitter.com/jawerty
LinkedIn: https://www.linkedin.com/in/jawerty/
Twitch: https://www.twitch.tv/jaredthecoder10x
Rumble: https://rumble.com/c/c-3572412

00:00 Introduction / What we're building
08:20 Learning about diarization / more planning
12:14 Setting up the colab
16:30 Getting Patrice O'Neal videos
22:35 Figuring out audio diarization
35:50 Torchtext versioning issue??
56:18 Looking for diarization alternatives...
1:06:28 Pyannote is finally working
1:10:40 Diarizing the audio files (segmenting based on speakers)
1:38:05 Making the code more robust
1:48:00 Implementing the audio segmenter (first draft)
1:53:58 Subscribe to Schemeology Podcast (my new podcast)
1:57:12 Back to coding
2:06:55 Trying a diarization alternative again...
2:30:15 It's not better
2:52:20 Let's just write out own audio segmenter
2:58:02 Resetting internet
3:01:25 We're back
3:07:07 Fixing things up
3:12:30 Writing the audio segmenter / dataset builder
3:34:42 Finally building out BASED dataset
3:36:43 Dealing with our chat troll / fixing our data
3:46:40 Chat troll asks for life advice / how to avoid bad influences in hs
3:55:18 Data cleaning / talk being surrounded by people who support you
4:05:55 Finally can start fine-tuning the model (llama 2)
4:14:25 Starting the fine-tuning / hanging with chat
4:30:40 Restarting fine-tuning and talk about getting involved in school
4:41:06 A Little LLM Fine-tuning PRIMER
4:51:50 Model is learning...we're waiting
5:16:53 Testing the Tate trained model
5:21:10 What to do to make it better
5:25:14 Getting more data
5:27:27 Using Zherka interviews
5:40:20 Parsing the audio segments for zherka data / Talk with chat
6:07:55 Running the training loop again
6:15:20 Waiting / Listening to Lofi / Talking about life
6:45:48 Trying out BasedGPT again
6:55:55 It fucking works haha
6:59:50 Saying goodbye