v0.0.5

Added

• textstream project: Enabled training of the GPT model using text streams in either a browser or a Node.js environment (eliminating the need to load the entire dataset into memory)
• In the train method, both maxIter and epochs parameters are now supported. Switched from df.forEachAsync to while loop + ds.iterator() (it was not clear how to terminate long-running forEachAsync).

Fixed

• Resolved memory leakage issue occurring in the generate method
• Use config.lr in the Adam optimizer