First release of the library

What's Changed

• feat: add tokenizer abstraction by @ayeganov in #1
• Show mean and std properly separated by @dariocazzani in #3
• Move model to device by @dariocazzani in #2
• Count flops and params by @dariocazzani in #4
• feat: Big refactor by @ayeganov in #5
• ⚙️ Fix experiment config loading + pathlib migration by @dariocazzani in #6
• 🛠️ Add code quality tools and CI linting workflow by @dariocazzani in #7
• update readme by @ayeganov in #8
• move loss out of the model by @ayeganov in #9
• Feat/large file handling by @ayeganov in #10
• feat: abstract away tokenizer as a json file by @ayeganov in #12
• Fixed roadmap formating in README by @dariocazzani in #13
• Create LICENSE by @dariocazzani in #14
• Refactor: Decouple Core Logic into a Reusable Library by @ayeganov in #15
• Prepare project for PyPI packaging by @google-labs-jules[bot] in #16
• clean up of unnecessary stuff by @ayeganov in #17
• 📚 Add examples/simple.py: minimal training demo by @dariocazzani in #18
• Chore/packaging cleanup by @ayeganov in #19
• Bootstrap HF Datasets by @ayeganov in #20
• 🪟 Add sliding window iteration mode for datasets by @dariocazzani in #21
• Started a fringe.md chain of thoughts by @dariocazzani in #22
• Use chunking for speed by @dariocazzani in #23

New Contributors

• @ayeganov made their first contribution in #1
• @dariocazzani made their first contribution in #3
• @google-labs-jules[bot] made their first contribution in #16

Full Changelog: https://github.com/LabStrangeLoop/scratchgpt/commits/0.5.0