Discord?

[NumseBacon]

Yo can I contact you on discord

[AndrejGobeX]

Sure, how can i send you a dm here :) ? Or better ask here so others can see, it might be of use of other people

[NumseBacon]

I have no idea how to dm o.o

[AndrejGobeX]

Post it here

[NumseBacon]

Well I just need help with tensorflow lol :p

[AndrejGobeX]

Yeah, I really should put Setup section in readme What is the issue?

[NumseBacon]

https://gyazo.com/24f5c476f3ea4746f3d0b781c56a9915 this :/

[AndrejGobeX]

I do not intend to open random links, sry Could you say what seems to be the problem

[NumseBacon]

[AndrejGobeX]

Hmm, haven't had such problem, r u on windows?

[NumseBacon]

Yup

[AndrejGobeX]

I can only think of a dependency missmatch. Either try to find on stackoverflow or try to create a new virtual env for python (or just remove it and reinstall), then just install tensorflow I know it's not the answer you were looking for but i really don't know Python 3.9? Maybe try 3.7 or 8

[NumseBacon]

What is your tensorflow version?

[AndrejGobeX]

Don't remember, removed it recently

[NumseBacon]

Dang. Ive already tried all of those things multiple times :/

[AndrejGobeX]

Tensorflow has dependencies Install only tensorflow with python 3.7 or 8 If that doesn't work, try to install a previous version of tf I haven't used cuda etc

[NumseBacon]

could it be because im using tf v2?

[AndrejGobeX]

No, i used v2

[NumseBacon]

I just installed a new version of tenserflow (2.2.0) and it installed the dependencies but still shows that error

[AndrejGobeX]

Try an older version, or older python, or research the error

[NumseBacon]

Yea ive searched a lot and havent found anything

[NumseBacon]

Ok so, I managed to do everything but I don't know what to do lol I ran trainer.py but idk

[NumseBacon]

Im trying to run the screenrecorder but when I do it saus not enough arguments. I figured out what PID was but I have no idea what speed address and endian is :/

[AndrejGobeX]

I will try to add everything in readme For now, speed address is the memory address of the speed variable in trackmania (you use CheatEngine) PID is the ID of the trackmania (open task manager and find the PID) when running scripts, at the top of the script is usually a comment telling you how to run a script You can explore them for yourself Screenrecorder is used to take snapshots while you play the game. These are used to train the nets For now, try driver.py Run it like: python driver.py net pid speedaddress (net should be either Neptune or Maniack) there is an example in the script run trackmania and then run the script, once you see ready or press s to start, alt tab to trackmania (you should already be in the driver mode), press delete to respawn and hit s to start the ai endian dont worry it all set up

[NumseBacon]

How do I find the speed variable in TM? Under memory view or?

[AndrejGobeX]

CheatEngine Run it, select trackmania, and find the value There is a simple tutorial for it

[NumseBacon]

well im pausing the game and putting the speed value into the cheat engine search and doing it multiple times, but I end up with values that stop working after ive completed a track

[AndrejGobeX]

yeah thats a problem i know, when I use this, I usually respawn before the end When u finish everything resets

[NumseBacon]

just to be sure, what value type?

[AndrejGobeX]

default, im not sure what u mean

[NumseBacon]

In cheat engine

[AndrejGobeX]

just leave default

[NumseBacon]

none of these are working

[NumseBacon]

as in it gives me not enough arguments in cmd

[AndrejGobeX]

depends on the script open driver, look at the top there is an example fill in all the stuff (place the AF19C... for speed) should run

[NumseBacon]

yea this and still not enough arguments

[AndrejGobeX]

call python python Driver....

[NumseBacon]

thats stupid

[AndrejGobeX]

it works?

[NumseBacon]

Well yea. I wait for it to say ready, and then I click s and it waits like 20 sec before anything happens and then I try to click q and it doesnt stop

[AndrejGobeX]

hmmm, i dont now the specs of your pc, but i can tell you that you can tap q multiple times or just alt tab to cmd and ctrl+c couple of times (not the best solution, but it aint the point of the project) as for the delay, dont really know, press s few times or hold it longer

[NumseBacon]

Ok so! Its driving kinda. When the count down reaches 0 it takes a few seconds for it to drive, SO now do I kinda train it do become faster

[AndrejGobeX]

now u can do whatever you want, explore the repo and enjoy

[NumseBacon]

But how do I train it to become better

[AndrejGobeX]

trainer + screenrecorder explore the repo on your own, its easy to understand

[NumseBacon]

So just drive a lot with screenrecorder and then trainer.py?

[AndrejGobeX]

yep, but machine learning is much more than that just research play etc

[NumseBacon]

idk what you mean by research play

[NumseBacon]

You also said "I plan to do reinforcement/evolution/mutation net in the future" when u gonna do that?

[AndrejGobeX]

So I tried it with the latest versions of everything and it worked well, I fixed a small issue considering speedcapture It sould work fine now As for RL, I'm working on it