Corpus2GPT v0.1

Highlights

• Synthetic language translation dataset generator added.
• Fixed the issues relating to import directories.
• Removed all double importing bugs in the code
• Removed all useless docs from the code.

Corpus2GPT v0.1.0-alpha

Corpus2GPT - The easiest way to make and train GPTs from scratch.

Key Features

• Classical Multihead Attention
• Decoder Module
• Random Sampling Search Strategies
• Multiple Language Support
• Sentence Piece Tokenizer and Vectorizer
• GPT Builder
• Distributed Learning Support
• Scaling Utilities Support

Currently, in alpha phase, serialization and de-serialization will be implemented in beta version and import directories would be changed to support import Corpus2GPT.

Full Changelog: https://github.com/abhaskumarsinha/Corpus2GPT/commits/v0.1.0-alpha