Video wird geladen...
Video konnte nicht geladen werden
Releasing PufferLib 4.0: Train agents in seconds
216,829 Aufrufe โข vor 6 Monaten โขvia X (Twitter)
39 Kommentare
congrats! ๐ฅ hit me up if you want to hop on stream next week to showcase the new version!
Can this be setup with an autoresearch type loop to train for a task?
no link? seriously?
Algo punishes direct links
Assuming this is true, which I doubt, you prefer to boost you tweet than to lead people to your site?
Puffer is insanely fast now ๐
nice!
nice i have some interesting experiments with it for stock trading, repo is balooning out with lots of slop though due me running autoresearch agents rn also moving to c for bitbankNZ my forecasting platform but taking a bit of time lol
very cool
Congrats! Trailer looks nice (is that raw raylib?!)
raylib + shaders. Trailer guy was out so I just did it
@max_paperclips Banger
nice!! pufferlib rules!
Awesome work and huge congrats!!
examples don't work on this branch
which examples?
lets goo :) super excited! ๐ก
If I update to 4.0, do I need to change anything in my workflow, or is everything already compatible with previous 3.0 pipelines, including the environment, model, .ini, and so on?
Incredible
jaw dropped on those data vizs, i was not fair to you the last time i left a comment! i'm eating my words rn
Congratulations on pushing this out! Incredible work, 20M+ SPS is wild.
@1million_ae have a look
Feed the puffer joey
Well done!
amazing
Ah superb.
Awesome
Yeah, this is wild!
wdym agents?
๐ฅ
Congratulations, Joseph, thank you so much for your amazing work! <3
PufferLib 4.0 trains agents in seconds? Cute, prove it or I'm calling shenanigans, bruh.
Congratulations man!๐ซก
feels like this only works in theory, what about when the data gets complex? learned this the hard way
Wao
gg well played ! congratulations again. 4.0 is absolutely nuts
Is it possible to help explain how I can use this to perform reinforcement learning on trading? Like Iโm sure thereโs simple strategies (and therefore less variables to test, I expect) that could be used to significantly increase the quality of trading signals.
congrats joseph,
Amazing !
