Now, when you specify a separate folder for the model, the “opensoftware_world_osw1_tokenizer.model” file will also be searched for in that folder.

The model training and model initialization code has been updated. (CPU/GPU auto-detection has been added.)

We have switched from the Regex tokenizer to the SentencePiece tokenizer. (The updated code was written by OpenSoftware-World and corrected by ChatGPT.)

Initialization code for the OpenSoftware-World-OSW1 AI model. (This code was written by Claude and edited by OpenSoftware-World.)