2026.0123.2251
2026-01-25
the "PyRun_StringFlags() returns Nil
but PyErr_Print() outputs Empty" Problem Solved and the "opencv2.a" Removed Version;
we Changed Three Things at this time.
****1) we Solved the the "PyRun_StringFlags() returns Nil
but PyErr_Print() outputs Empty" Problem;
When you go to RunScriptView for the first time after you installed this App,
you needed to once go back to EditScriptView and then needed to go to RunScriptView again.
To Solve the "PyRun_StringFlags() returns Nil
but PyErr_Print() outputs Empty" Problem,
it was needed to once call Py_FinalizeEx() and then to call Py_Initialize() again,
Means it was needed to Re-initialize the python interpreter.
It was Hard to find this solution
'cause the error message "fopen failed for data file: errno = 2 (No such file or directory), Errors found! Invalidating cache..."
has Hidden this problem for many years.
As a result, the error message "fopen failed for data file: errno = 2 (No such file or directory), Errors found! Invalidating cache..."
didNot relate to
the "PyRun_StringFlags() returns Nil
but PyErr_Print() outputs Empty" Problem.
****2) we Removed "opencv2.a";
'cause you canNot use opencv Python Bindings Nor opencv C++ Bindings
on 'PeterJohnNumPyConnectFour'.
By removing"opencv2.a",
you Can Save 180MBytes of SSD Disk Space and downloading Time.
Best Regards,
--Yasushi Yassun Obata
2025.0927.1334
2025-09-29
"s003ConnectFourMctsCppGame.py" and "s004ConnectFourTraining.py" BugFix Version;
We Apologize that
MCTS, MonteCarloTreeSearch, C++ Version had a Bug that MCTS C++ version did Not explore the Game Tree,
so it was Very Weak.
We Fixed the Bug. Now MCTS C++ version explores the Game Tree.
So MCTS C++ Version Became Strong.
****TutorialSeason005/s003ConnectFourMctsCppGame.py
Became Strong.
****TutorialSeason005/s004ConnectFourTraining.py
which Uses the MCTS C++ Version to Train.
in The Results of Training,
RedReward and YellowReward became Almost Same Values such as 64.5 versus 64.
it means the MCTS C++ Version became Strong.
****TutorialSeason005/s005ConnectFourGame.py
which Uses the Trained Data on "s004ConnectFourTraining.py".
so it became Strong.
****Why did we made MCTS, MonteCarloTreeSearch, C++ Version??;;
NumPy is Ten times Faster than Python,
and C++ is Ten times Faster than NumPy,
As a Result, C++ is Hundred times Faster than Python.
Enjoy Deep Learning.
Enjoy Reinforcement Learning.
Best Regards,
--Yasushi Yassun Obata