v0.9.7 Major source code update release

Source release only.
The complete source code of the GUI has been ported back from CodeTyphon to Lazarus IDE with a lot of effort.
Reason for this was the possibility to do an automated full compilation and building process in virtual machine environments right on Github.
Github workflows and actions are set up and allow that now. Code is checked for successful build process and compilation in virtual machines automatically.
The code was cleaned up, several changes were made to make it work.
For now, the code for the GUI builds fine now in Lazarus stable/latest per OS versions (2.2.2) - on Windows 10, Ubuntu Linux 22.04 and MacOS (latest) virtual machines with all prerequisites automatically installed and configured to make it happen. A goal is to have a fully automated build and publishing process setup, no local build chain needed anymore!
The source code of the GUI was moved into a new folder "src" to keep the main directory for executable and future configuration files only.
There is a folder with a standard i8n language file. So localization is possible and will happen at a later point in time, will do the German localization myself.
Next efforts will go into
a.) ...do shell versions of the scripts OR do a direct process start of the Python scripts from the Pascal main process.
b.) ...making the shell scripts/process starting instructions work in Linux and MacOS.
c.) ...extensive testing of the underlying for better understanding on my side for the Python based training code. I noticed problems getting results for a parameterized model. Non-parameterized models trained fine (and fast!) though.
d.) ...investigating the possibilities to stream stdio back into the GUI and run the Python scripts silently to keep users from interacting with separate windows directly. Inter-process communication is not easy though, so far I did not get satisfying results in interfacing Pascal and Python directly. Python4Lazarus/Delphi does not integrate well for me so far.
e.) ...making QOL changes to the GUI, making use of MessageBoxes to give feedback to the user about successful training process or missing data etc...

This is an intermediate (huge) source code only update to make maintenance of the project more future proof, OS independent and user-friendly in the end.

v0.9.6 Major GUI re-arrangement, localization preview, minor fixes

This one includes an executable binary installer (portable) again.
The GUI now reflects the workflow much better, left to right, top to bottom.

HAVE FUN.

v0.9.5: Merge pull request #32 from smallbutfine/smallbutfine-patch-6

Batch scripts updated and fixed

Just overwrite installed files.

v0.9.4 Now uses nightly version of PyTorch 2.0 for great performance gain.

Same as previous release, but now uses bleeding edge PyTorch package (2.0), which needed also latest SciPy.
Great performance gain.
Changed GUI for better understanding of the building process.
Model installation button should now be functional.

Binary installer for CPU only model training, now uses latest SciPy and PyTorch packages for best performance.
USE THIS ONE IF YOU HAVE A WEAK GPU OR ENCOUNTER PROBLEMS WITH GPU BASED TRAINING.

Binary installer for GPU based model training, now uses latest SciPy and PyTorch packages for best performance.
USE THIS ONE ONLY IF YOU HAVE AN NVIDIA CARD WITH ENOUGH RAM. IF IN DOUBT, USE THE CPU ONLY INSTALLER.

(Next release will probably use the next level of PyTorch technology introduced with 2.0 for even better performance, I am investigating the changes needed in the training codebase.)
Beta state.

v0.9.3 Now with offline local GPU model training support, approx. 20x faster than CPU training.

Finally, I found a PyTorch and CUDA version, that did work on my Nvidia RTX 3080 card with 10GB RAM. Win 10.
This accelerates the model training times drastically, even compared to CPU rendering on an AMD Ryzen 5950X.

DO NOT INSTALL THIS VERSION IF YOU HAVE A SLOW GRAPHICS CARD OR LOW VIDEO RAM from Nvidia.
You will almost certainly run into problems while training the models.

A graphics card with above or similar specs to mine (Nvidia), or at least enough RAM, might run fine.

Stick with the previous release, if in doubt.

Installation:
Unpack the loose files in the source code archive below over the installed files of the binary release v0.9.0 (portable install) or the binary release v0.9.0 (portable install) + v0.9.2.
(If there is no overwrite warning to confirm, you did something wrong...)
Next release will contain a full installer again.

Start the .exe file of the portable release and re-/install the Build System from the GUI.
Have fun.

v0.9.2

small changes, untested

This should work for now.
Just drop the loose files inside the folder over the previous installation, overwrite, and the problem with parameterized model training should be gone.

v0.9.0

Initial release

Install/Extract to any folder you have access rights, e.g. documents, desktop or even a thumbdrive.
Run the EXE file in the extracted folder.
That's it.

Have fun.

Initial release means: There CAN be errors or missing functionality. Still, this made my life so much easier...