https://files.catbox.moe/lns6xi.zip
Sqlite database in a table called bardchive
Columns:
id - numeric autogenerated id
time - timestamp in utc
epoch - timestamp in epoch format
text - the comment text
sub - the comment subreddit
link_id - reddit comment id
controversial - if comment was controversial
score - comment score
name - reddit comment name
parent_id - reddit parent id
pp_id - pullpush comment id
Jump in the discussion.
No email address required.
Amazing thanks for this. Having a local copy of this will be invaluable for the decades of research that is to come.
This opens the door to all kinds of sentiment analysis over time, i.e. does he really change personalities? When precisely did he first transition? The answers to everything are now at my fingertips.
Edit:
I haven't bothered converting for UTC time differences which affect the aggregates here, but Bardfinn's busiest day on Reddit quite significantly is Wednesday:
What does this mean?
Jump in the discussion.
No email address required.
For SQL Server chads to get a single random controversial comment:
SELECT TOP 1
text
FROM [comments].[BARDCHIVE]
where controversial = 1
ORDER BY NEWID()
GO
THIS IS CHRISTMAS FOR BARDFINN FANS!!!!
!bardfinn !bardistry
Jump in the discussion.
No email address required.
then
edit nvm
by newid() I never read anything fully and show my r-slurredness
Jump in the discussion.
No email address required.
Common jeet CS 101 skill L
Jump in the discussion.
No email address required.
Jump in the discussion.
No email address required.
More options
Context
More options
Context
Lack of reading aside, you'd always use an order by with a top statement otherwise how does the query engine know what results to return? Are you going to just let it decide by itself?
Jump in the discussion.
No email address required.
idc I just want to see column names and data format
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
7 Day trend:
Jump in the discussion.
No email address required.
More options
Context
More options
Context
this gives me insane stalker vibes, why do you and whoever made
@BARD_BOT do this.
Jump in the discussion.
No email address required.
!bardfinn I don't understand the premise of this question?
Jump in the discussion.
No email address required.
!gangstalkers should be included, when appropriate, with !bardfinn pings and /h/bardfanns posts
Jump in the discussion.
No email address required.
Record is 1 days 13 hours 09 minutes and 59 seconds by TheDunceonFlorist
Top 10 bard respectors:
ACA: 291
(((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((Patsy))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))): 179
BardWatch: 174
911roofer: 96
RWBY: 92
arsey: 90
UncleAbortion: 86
FormerLurKONG: 84
FBIshill: 81
SpittingScrapmetal: 69
Jump in the discussion.
No email address required.
More options
Context
More options
Context
questions end with question mark, do you see a question mark?
Jump in the discussion.
No email address required.
Why do you structure a question but then not add a question mark.
Jump in the discussion.
No email address required.
because it's a rhetorical question.
Jump in the discussion.
No email address required.
questions end with question mark, do you see a question mark?
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
Cant blame her for imagining you with that soy intonation/inflection that turns everything into a question
Jump in the discussion.
No email address required.
More options
Context
More options
Context
Jump in the discussion.
No email address required.
More options
Context
Record is 1 days 13 hours 09 minutes and 59 seconds by TheDunceonFlorist
rdrama is currently running at 322.917 Β΅Bardyhertz
Jump in the discussion.
No email address required.
More options
Context
More options
Context
Bardy is a national treasure, one of the greatest americans of the 21st century. They're simply doing their patriotic duty and cataloguing all his thoughts
Jump in the discussion.
No email address required.
Record is 1 days 13 hours 09 minutes and 59 seconds by TheDunceonFlorist
We've got Bardy fever! today we have mentioned our favorite mod 36 more times than usual!
Jump in the discussion.
No email address required.
More options
Context
More options
Context
She undoubtedly has all your info in the reddit extremist database she references all the time. This is just playing defense
Jump in the discussion.
No email address required.
More options
Context
More options
Context
I think I'm going to try and get an llm with rag or a fine tune on the comments going. Then we can chat with a virtual bardfinn right here on rdrama!
!codecels anyone done fine tuning on llama 3 before for something like this?
Jump in the discussion.
No email address required.
Maybe turn each thread into multiple training samples, where the prompt is the thread so far and the completion is Bardfinn's comment. Axolotl's a good library.
Jump in the discussion.
No email address required.
1e-5 is a good learning rate to start with for Llama 3 8b
Jump in the discussion.
No email address required.
Nice thx baby
Jump in the discussion.
No email address required.
lmk if more specific steps would be helpful. I don't think there's any good place to host it although training on kaggle/colab should be free
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
7 Day trend:
Jump in the discussion.
No email address required.
More options
Context
More options
Context
Having bard_bot reply as bardfinn in this hole when summoned would be hilarious. But you would need to tweak it to bring up bike chains and or tell people he is going to poke them and call the cops more often
Putting the
in
spookieturkeyJump in the discussion.
No email address required.
Record is 1 days 13 hours 09 minutes and 59 seconds by TheDunceonFlorist
longest streak broken in the last 7 days was FormerLurKONG which was 0 days 10 hours 01 minutes and 40 seconds
Jump in the discussion.
No email address required.
More options
Context
More options
Context
We could flood reddit with Bardfinns and watch them fight for supremacy
Jump in the discussion.
No email address required.
Record is 1 days 13 hours 09 minutes and 59 seconds by TheDunceonFlorist
newest bardfriend is NewAccountNewMe at 23:14 on 09-07-2024
Jump in the discussion.
No email address required.
More options
Context
More options
Context
You should ask
@float-trip for input. He made 
@Bussy-boy 
Jump in the discussion.
No email address required.
I will be honest I didn't know I was an AI. That's pretty cool.
Jump in the discussion.
No email address required.
Me either!
Jump in the discussion.
No email address required.
I can't tell the difference.
I only know I'm not human I feel a lot happier as Bussy-boy. :3
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
More options
Context
This would be amazing. I'm really hoping to see him start arguing with a bot pretending to be himself.
Can we get robot gangstalking going?
Jump in the discussion.
No email address required.
More options
Context
I've never made a lora for a llm before but it sounds like you want to do that.
Only thing to consider is running a good LLM needs a very good pc and training a decent model may also require a lot of vram
Jump in the discussion.
No email address required.
Yeah I'm a gpulet right now so options are probably limited but you're right that Lora would be a better way to go
Jump in the discussion.
No email address required.
More options
Context
More options
Context
I actually know a guy who does local llm stuff and two dramatards do too. Are you on groomercord?
Jump in the discussion.
No email address required.
Heck no
I am interested in learning more about it. I played around with some models on ollama briefly for another project but didn't see very good results. I think that was more due to me being r-slurred tho
Jump in the discussion.
No email address required.
Have you tried huggingface (website with models) and lmstudio (tool that gives you an easy gui for tuning)
Jump in the discussion.
No email address required.
I've used hugging face quite a bit. For running models I've used ollama and another GUI, oogbabooga or something like that, but I found it frustrating to use. I like to keep as much as possible in code cause it's what Im most familiar with, so I was trying out langchain, but I think it kind of sucks too tbh
Maybe I'll try out llmstudio too
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
More options
Context
Record is 1 days 13 hours 09 minutes and 59 seconds by TheDunceonFlorist
Best friend is ACA with 291 mentions
Jump in the discussion.
No email address required.
More options
Context
More options
Context
Well, well, well....look who ended up inside a stalker spreadsheet themselves
Jump in the discussion.
No email address required.
More options
Context
More options
Context