Testing Neural Network – Python AI in StarCraft II tutorial p.12




[ad_1]

In this part 12 of the artificial intelligence in StarCraft II with python series, we’re going to cover the code used to actually test the model in game, and some of the results I found.

Text tutorials and sample code: https://pythonprogramming.net/using-neural-network-starcraft-ii-ai-python-sc2-tutorial/

Chat with us on Discord: https://goo.gl/Q9euv3
Support the content: https://pythonprogramming.net/support-donate/
Twitter: https://twitter.com/sentdex
Facebook: https://www.facebook.com/pythonprogramming.net/
Twitch: https://www.twitch.tv/sentdex
G+: https://plus.google.com/+sentdex

Source


[ad_2]

Comment List

  • sentdex
    December 2, 2020

    The AI doesnt really work! it only will learn that always using choise 3 is "best". U dont have to get 2.8k wins for that 200 is enough to do that…

  • sentdex
    December 2, 2020

    from sc2.constants import NEXUS, PROBE, PYLON, ASSIMILATOR, GATEWAY, CYBERNETICSCORE, STARGATE, VOIDRAY, SCV, DRONE, ROBOTICSFACILITY, OBSERVER .. is not working

  • sentdex
    December 2, 2020

    Great video series! I think this could help get a more intuitive feel for how machine learning works. The only problem I'm having is that you seem to tell us what you're going to do a lot. Brief overviews are fine now and then, but I'd rather just see you do it, and explain it as you do it. Also, I'm taking back my previous vote on a different video when you asked if we'd like copy-paste + explain, or to see you debug. If it's basic syntax, then of course I'd rather copy+paste, but if it's an error that is logic or missing a header, etc., I think it's helpful to see you go through the process of fixing that. I watch videos to learn from other people's mistakes too. Thanks again for doing all these series! You got me going on Python in 2 days. That's impressive!

  • sentdex
    December 2, 2020

    After changing just one fucking value… Keras works now for me xD (after 10 hours now)

  • sentdex
    December 2, 2020

    My AI Bot won't stop bragging about his associates degree on PUBG

  • sentdex
    December 2, 2020

    You deserve more views and more subscribers. Always loves your content.

  • sentdex
    December 2, 2020

    Could you please make a video on your full sublime setup? So settings, extensions and stuff?

  • sentdex
    December 2, 2020

    Cool bot but can it beat the cheesers on the NA ladder?

  • sentdex
    December 2, 2020

    I'm hyped for part 13!

  • sentdex
    December 2, 2020

    Pretty good. 👍👍👍

  • sentdex
    December 2, 2020

    Hi sir please i need to download image with python and selenium
    Not urllib or requests or beautifulsoup thanks

  • sentdex
    December 2, 2020

    hi master sentdex, just passing by to check the next tutorial part… im anxious for it

  • sentdex
    December 2, 2020

    Hi sentdex, i see your video "Haar Cascade Object Detection Face & Eye – OpenCV with Python for Image and Video Analysis 16" and i want to ask you something like how i can add images like glasses (like snapchat) and how i can add take photo like phone (like snapchat)

  • sentdex
    December 2, 2020

    my dear friend, you are a great proffessor, please continue making this awesome series. You the best and you never play yourself .

  • sentdex
    December 2, 2020

    Need more of this please

  • sentdex
    December 2, 2020

    Hi, what about this? UserWarning: Error in loading the saved optimizer state. As a result, your model is starting with a freshly initialized optimizer.
    warnings.warn('Error in loading the saved optimizer '

    Thank you in advance!

  • sentdex
    December 2, 2020

    You know so much, yet so humble. Thank you.

  • sentdex
    December 2, 2020

    What do you get if you sponsor ?

  • sentdex
    December 2, 2020

    Hey sentdex, love this series!
    A quick question – don't you think that giving the "do nothing" choice of action a consistent time period, would benefit the algrothim? Both in training and prefromance, I think that a more predictable action lets the bot a much better chance of learning to use it right. Ideally I would set it to be just a second or two, not to get stuck at it for too long.

  • sentdex
    December 2, 2020

    Keras optimiser started randomly crashing the game for me, sad times

  • sentdex
    December 2, 2020

    You should look at professional tournament of StarCraft 2.

  • sentdex
    December 2, 2020

    What about micro? A machine learning implementation to optimize a 6 vs 6 stalker micro battle with or without blink sounds interesting. What about letting the ai decide a gateway unit composition (what to train next and the number of gateways)?

  • sentdex
    December 2, 2020

    Upvoted for verbal AI encouragement

  • sentdex
    December 2, 2020

    nice regular cup

    ik this is off topic , but im wondering an AI if can be made to have free will
    P.S.: i start to think you drink off cups in the video just fro me to say "nice cup"

  • sentdex
    December 2, 2020

    Glad to know that the model works. I finally managed to read (which you provided) and feed the data into the Keras model using the tf dataset API and started training on my laptop. I will share my code if anybody is interested. Also another thing I noticed was that you don't seem to normalizing the xdata (image) by dividing it by 255 before feeding it to the CNN (unless I missed this part). Wouldn't this negatively affect the training?

    A few things which I noticed could be improved:
    1) The x data (image) is being stored as an uint8 datatype. But the y data (one hot array) is being stored as float64 datatype. I think changing it to Bool datatype could reduce size of the .npy files.
    2) The logic for issuing order to create buildings and units could me made into a general method. It could take as argument the type of building/unit to be created and it's per-requisites.

  • sentdex
    December 2, 2020

    Is it possible to give the neural network full control of everything – like the what and when to load troops, how to command troops, when and which buildings to make and other similar commands? I'm kinda new to neural networks so I don't know if this is a weird question or not, but I thought it to be the best course of action as it would result in the NN having more control and therefore allowing it perform better right?

  • sentdex
    December 2, 2020

    Thank you for being the minority of youtubers who actually do computer education so well.

Write a comment