Setting up ML Agents for Unity in 5 Minutes 

This has been bothering me for a while too. It's upsetting because I cannot reproduce this behavior. You don't need any C++ tools because pip builds the binary version of numpy. If you want to manually install it from source code, then you would need C++, but I have never done that. One thing I can suggest, is running this *pip install --no-cache-dir setuptools wheel* This installs *steuptools* and *wheel* libraries. Try installing numpy after that (make sure you are still in the venv). Not sure if it will fix you issue, but its worth a shot As for following another tutorial, you definitely can try. However, I made this tutorial because following them came with a lot of headaches such as conflicting library versions, outdated python versions, etc. Most tutorials cover the same thing anyway. If you find a solution, please let me know about it so I can pin the comment!

@@pingcode I tried installing the setuptools and wheel libraries as you suggested, but it didn't work. The good news is that I found a comment on CodeMonkey's video suggesting a similar procedure with different versions of Python and the Unity editor. It worked for me, and the "mlagents-learn --help" command worked just fine. I'll juste paste the comment right here : " Hi from 2023! I had some trouble instaling the python enviroment but I made it work. Unity: 2022.3.14f1 Python 3.9.13 Before venv: py -3.9 -m venv venv After venv: python -m pip install --upgrade pip pip install mlagents pip3 install torch torchvision torchaudio pip install protobuf==3.20.3 pip install packaging Works fine! Hope so u too :D " I started from scratch, created a new project with the suggested Unity version, and it worked. #### Another question: Does it matter if I install the ML Agents package (from the Unity editor) after the whole CMD procedure? Will it impact my simulations? Thank you for the quick response btw.

@@IsmaiArrouchdi No it shouldnt make a difference. You might have to restart Unity or CMD if it doesn't detect its been installed yet. But that's it. So looks like using a different python version helped you, that's good to know. I will pint the comment for others to look at. Thanks!

Glad it helped!

Appreciate it! Good luck with your project

Glad I could help!

You're welcome! 😊

I'm so glad it helped!

Thank you 😊

Glad I could help!

No problem!

I am not sure, I never tried it. Please look at these links, they might be helpful in some way. forum.unity.com/threads/gpu-training-using-ml-agents-cuda-cudnn.910652/ www.tensorflow.org/install/pip github.com/miyamotok0105/unity-ml-agents/blob/master/docs/Installation-Windows.md www.reddit.com/r/Unity2D/comments/16nko4k/how_to_utilize_cuda_during_the_training_of/

I'm assuming you ran the tensorflow command in the terminal? I can't help you without more information. A common mistake is that a training session didn't finish yet, so no grpah can be made (which is indicated by the number of `steps`). Let me know what you have tried and I'll try to help you. In the meantime, make sure you read this unity-technologies.github.io/ml-agents/Using-Tensorboard/

@@pingcode Yes looks like I excited it too early. Thank You

Honestly no idea since I never used the agent asset. I'm assuming you mean the little blue character? I've never used it so I don't know. If that's not what you mean, can you be a bit more specific

@@pingcode yeah it was and actually i resolved it by just using the latest mlagent version tar.gz and extract it and used its Project folder to access the assets properly (including the steps you shown in this vid)

@potato8938 That's great to hear 👍

@@KhaledHossamShosha Awesome!

@@pingcode Dose i need a CUDA in this release ?

@KhaledHossamShosha do you have an nvidia gpu? Cuda is just used to speed up training to my understanding. I don't know much about it. You don't NEED it

@@pingcode i have RTX3050

@@KhaledHossamShosha yeah so then you should be able to use it. Try look up some documentation or videos

It's hard to say without knowing what you did. The error message is saying that you are trying to import numpy from its source directory. You might have done something wrong when changing directories in the terminal or you might have numpy installed in a weird way on your machine. Is there any part of the tutorial that you are confused about and need help with?

Did you make a virtual environment? If not, you might be using a version of numpy already installed on your pc. If you did make a virtual environment, try to unistall numpy and reinstall it from within it The error might be indicating a version of numpy not compatible with the current version of python or mlagents.

@@pingcode got the same error when trying to install numpy 1.21.2 on the newest version it just doesnt even tries to create that "wheel" yes i am in an virtual environment heres an List of all the installed packages(if it helps): Package Version ----------------- --------- filelock 3.15.1 fsspec 2024.6.0 intel-openmp 2021.4.0 Jinja2 3.1.4 MarkupSafe 2.1.5 mkl 2021.4.0 mpmath 1.3.0 networkx 3.3 numpy 1.26.4 pillow 10.3.0 pip 24.0 setuptools 57.4.0 sympy 1.12.1 tbb 2021.12.0 torch 2.3.1 torchaudio 2.3.1 torchvision 0.18.1 typing_extensions 4.12.2

@@Phoenix_3223 Thanks for that. Upon checking my pip install list, I see I have a lot more packages installed. Please check this link to download my requirements.txt file I just made: www.dropbox.com/scl/fi/phe813xzccz69riaf678x/requirements.txt?rlkey=etb53insp2txwtz8x0yaqwmqx&st=nxd615dq&dl=0 You can try uninstall everything in your virtual environment and installing everything in this file using *pip install -r requirements.txt* Let me know if that works!

@@pingcode i did exactly that untill the installing just stopped because of the same error: ERROR: Could not build wheels for numpy, which is required to install pyproject.toml-based projects

I had this and had the microsoft store version which was 3.10.11 and installing 3.10.0 fixed it

Are you working in a venv? Is pip up to date? Did you install packages when using python version 3.10.0 and not just switching python versions? If yes to all then I really don't know. I made a dropbox link to my requirements.txt file if you want to try install everything using that? www.dropbox.com/scl/fi/phe813xzccz69riaf678x/requirements.txt?rlkey=etb53insp2txwtz8x0yaqwmqx&st=vdsaergu&dl=0 Delete venv folder, and remake it. Put the file in the venv folder. Activate the venv and run *pip install -r requirements.txt*

@@pingcode Thank you this worked you have literally saved my dissertation

@@nicholaspaleologos7088 awesome, glad it worked out!

@iaxsgames6029 Are you working in a venv? Is pip up to date? Did you install packages when using python version 3.10.0 and not just switching python versions? If yes to all then I really don't know. I made a dropbox link to my requirements.txt file if you want to try install everything using that? www.dropbox.com/scl/fi/phe813xzccz69riaf678x/requirements.txt?rlkey=etb53insp2txwtz8x0yaqwmqx&st=vdsaergu&dl=0 Delete venv folder, and remake it. Put the file in the venv folder. Activate the venv and run *pip install -r requirements.txt*

@@pingcode Thank you- my pip wasn't up to date but know I have the error Descriptors cannot be created directly. (edit) I didn't read down, sorry

Can you please share more details as to what you are doing when you see the error? What command are you typing when you see this? What OS are you using?

What version of Python are you using?

@@pingcode sorry for the late reply, but i have 3.12.2

@@Spiex81 you need 3.10, check description

@@pingcode Ah ok thanks

Thank you

I had 2 versions of python installed on my computer. I needed to uninstall my recent version of Python to switch on the 3.10 version. I hope it helps someone ;)

@simonnaulet2118 you dont have to uninstall that python version! You can just switch python versions in the terminal when needed.

@@pingcode where terminal

opps nvm

Many people are having this issue and the reason isn't quite clear. The most I found online about this issue is that Numpy and Python are conflicting with each other. Try to install a different version of numpy. Make sure pip is updated. Make sure you are using Python 3.10.0 in the terminal.

@@pingcode that's probably the problem because when i try to install numpy i'm getting this error (and i'm using python 3.10 and pip is updated): (venv) C:\Users\yonat\ot>pip install numpy==1.19 Collecting numpy==1.19 Using cached numpy-1.19.0.zip (7.3 MB) Installing build dependencies ... done Getting requirements to build wheel ... done Preparing metadata (pyproject.toml) ... error error: subprocess-exited-with-error × Preparing metadata (pyproject.toml) did not run successfully. │ exit code: 1 ╰─> [90 lines of output] Running from numpy source directory. :460: UserWarning: Unrecognized setuptools command, proceeding with generating Cython sources and expanding templates C:\Users\yonat\AppData\Local\Temp\pip-install-ha3yov2t umpy_1fff654732424398a076fa9ec2c2648c\tools\cythonize.py:73: DeprecationWarning: distutils Version classes are deprecated. Use packaging.version instead. required_version = LooseVersion('0.29.14') C:\Users\yonat\AppData\Local\Temp\pip-install-ha3yov2t umpy_1fff654732424398a076fa9ec2c2648c\tools\cythonize.py:75: DeprecationWarning: distutils Version classes are deprecated. Use packaging.version instead. if LooseVersion(cython_version) < required_version: performance hint: _common.pyx:275:19: Exception check after calling 'random_func' will always require the GIL to be acquired. Declare 'random_func' as 'noexcept' if you control the definition and you're sure you don't want the function to raise exceptions. performance hint: _common.pyx:299:19: Exception check after calling 'random_func' will always require the GIL to be acquired. Declare 'random_func' as 'noexcept' if you control the definition and you're sure you don't want the function to raise exceptions. performance hint: _common.pyx:322:50: Exception check after calling 'random_func' will always require the GIL to be acquired. Declare 'random_func' as 'noexcept' if you control the definition and you're sure you don't want the function to raise exceptions. performance hint: _common.pyx:1002:32: Exception check after calling 'f1' will always require the GIL to be acquired. Declare 'f1' as 'noexcept' if you control the definition and you're sure you don't want the function to raise exceptions. performance hint: _generator.pyx:706:41: Exception check after calling '_shuffle_int' will always require the GIL to be acquired. Possible solutions: 1. Declare '_shuffle_int' as 'noexcept' if you control the definition and you're sure you don't want the function to raise exceptions. 2. Use an 'int' return type on '_shuffle_int' to allow an error code to be returned. performance hint: _generator.pyx:735:45: Exception check after calling '_shuffle_int' will always require the GIL to be acquired. Possible solutions: 1. Declare '_shuffle_int' as 'noexcept' if you control the definition and you're sure you don't want the function to raise exceptions. 2. Use an 'int' return type on '_shuffle_int' to allow an error code to be returned. Error compiling Cython file: ------------------------------------------------------------ ... for i in range(1, RK_STATE_LEN): self.rng_state.key[i] = val[i] self.rng_state.pos = i self._bitgen.state = &self.rng_state self._bitgen.next_uint64 = &mt19937_uint64 ^ ------------------------------------------------------------ _mt19937.pyx:138:35: Cannot assign type 'uint64_t (*)(void *) except? -1 nogil' to 'uint64_t (*)(void *) noexcept nogil'. Exception values are incompatible. Suggest adding 'noexcept' to the type of the value being assigned. Processing numpy/random\_bounded_integers.pxd.in Processing numpy/random\bit_generator.pyx Processing numpy/random\mtrand.pyx Processing numpy/random\_bounded_integers.pyx.in Processing numpy/random\_common.pyx Processing numpy/random\_generator.pyx Processing numpy/random\_mt19937.pyx Traceback (most recent call last): File "C:\Users\yonat\AppData\Local\Temp\pip-install-ha3yov2t umpy_1fff654732424398a076fa9ec2c2648c\tools\cythonize.py", line 235, in main() File "C:\Users\yonat\AppData\Local\Temp\pip-install-ha3yov2t umpy_1fff654732424398a076fa9ec2c2648c\tools\cythonize.py", line 231, in main find_process_files(root_dir) File "C:\Users\yonat\AppData\Local\Temp\pip-install-ha3yov2t umpy_1fff654732424398a076fa9ec2c2648c\tools\cythonize.py", line 222, in find_process_files process(root_dir, fromfile, tofile, function, hash_db) File "C:\Users\yonat\AppData\Local\Temp\pip-install-ha3yov2t umpy_1fff654732424398a076fa9ec2c2648c\tools\cythonize.py", line 188, in process processor_function(fromfile, tofile) File "C:\Users\yonat\AppData\Local\Temp\pip-install-ha3yov2t umpy_1fff654732424398a076fa9ec2c2648c\tools\cythonize.py", line 77, in process_pyx subprocess.check_call( File "C:\Users\yonat\AppData\Local\Programs\Python\Python310\lib\subprocess.py", line 369, in check_call raise CalledProcessError(retcode, cmd) subprocess.CalledProcessError: Command '['C:\\Users\\yonat\\ot\\venv\\Scripts\\python.exe', '-m', 'cython', '-3', '--fast-fail', '-o', '_mt19937.c', '_mt19937.pyx']' returned non-zero exit status 1. Cythonizing sources Traceback (most recent call last): File "C:\Users\yonat\ot\venv\lib\site-packages\pip\_vendor\pyproject_hooks\_in_process\_in_process.py", line 353, in main() File "C:\Users\yonat\ot\venv\lib\site-packages\pip\_vendor\pyproject_hooks\_in_process\_in_process.py", line 335, in main json_out['return_val'] = hook(**hook_input['kwargs']) File "C:\Users\yonat\ot\venv\lib\site-packages\pip\_vendor\pyproject_hooks\_in_process\_in_process.py", line 149, in prepare_metadata_for_build_wheel return hook(metadata_directory, config_settings) File "C:\Users\yonat\AppData\Local\Temp\pip-build-env-u1c354rr\overlay\Lib\site-packages\setuptools\build_meta.py", line 368, in prepare_metadata_for_build_wheel self.run_setup() File "C:\Users\yonat\AppData\Local\Temp\pip-build-env-u1c354rr\overlay\Lib\site-packages\setuptools\build_meta.py", line 497, in run_setup super().run_setup(setup_script=setup_script) File "C:\Users\yonat\AppData\Local\Temp\pip-build-env-u1c354rr\overlay\Lib\site-packages\setuptools\build_meta.py", line 313, in run_setup exec(code, locals()) File "", line 489, in File "", line 469, in setup_package File "", line 274, in generate_cython RuntimeError: Running cythonize failed! [end of output] note: This error originates from a subprocess, and is likely not a problem with pip. error: metadata-generation-failed × Encountered error while generating package metadata. ╰─> See above for output. note: This is an issue with the package mentioned above, not pip. hint: See above for details.

Yes, from my experience 3.10.12 is the latest version that works. This is odd because their documentation says 3.10.12 *or higher* unity-technologies.github.io/ml-agents/Installation/ If you don't want to get rid of your current python version, you can install python 3.10.0 and switch the version of python you use in the terminal when running the commands shown in the video

Ill give it a try and tell you, you deserve a sub , thanks man!@@pingcode

@@pingcode IT WORKED, THANKSSS you deserve a sub!!

@@pafodev No problem!

use python 3.9

@@deoz210 Does 3.9 work with ml-agents 21

apologies for the late reply, if you still have the issue. Do you know what numpy version you have? You might have an older numpy version installed on your PC that isn't compatible with python 3.10 This issue page might be helpful to read what others have said. Someone mentions at the end they used numpy version 1.23.3 and it works github.com/Unity-Technologies/ml-agents/issues/6008

@@pingcode I am running into the same issue. I am using Numpy 1.23.3 and python 3.10.0 and I also get the "ERROR: Could not build wheels for numpy, which is required to install pyproject.toml-based projects" message :/ Help would greatly be appreciated!

@@BastiaanNetjes I'm not sure since I never ran into this issue myself. The only thing I can suggest is to uninstall numpy and reinstall it to see if that fixes anything. To my knowledge, you might also need C++ Build Tools installed since python and some libraries like numpy use C/C++. Though this is speculation. Here is a link to the build tools: visualstudio.microsoft.com/visual-cpp-build-tools/

There are a lot of different libraries that need to work together in order to have ml agents run properly. Many of these libraries have version conflicts with different python versions. 3.10.0 is a version I found to be stable. 3.12 for example is a newer python version that ml agents doesn't support yet.

you could not what?

@@pingcode to install ml agents i made all like you

@@calalmuradov6162 well, what is the problem?

Are you using Python 3.10.0? I can't help you with a random error message.

@@pingcode Yeah i am using the python 3.10.