Skip to content

A straightforward method for training your LLM, from downloading data to generating text.

Commits on main

1a2ac8a
8c91fe8
style: fix lint findings across the repo
FareedKhan-dev committed 1 month ago
09fee65
fix(ui): job status and stop work on Windows and macOS
FareedKhan-dev committed 1 month ago
e6b68db
build: add py.typed markers
FareedKhan-dev committed 1 month ago
bc612a3
52aa613
fix(types): make the core packages pass mypy
FareedKhan-dev committed 1 month ago
a5304d9
fix(device): keep PyTorch's default CPU thread count
FareedKhan-dev committed 1 month ago
61b9fc9
feat(data): tiny datasets for training on a laptop
FareedKhan-dev committed 1 month ago
f0bb7c9
6323899
feat(sft): LoRA adapters written from scratch
FareedKhan-dev committed 1 month ago
31de9a9
feat(grpo): Dr. GRPO, DAPO and GSPO options
FareedKhan-dev committed 1 month ago
41af875
feat(dpo): IPO, SimPO and conservative DPO
FareedKhan-dev committed 1 month ago
aabecd2
feat(optim): Muon optimizer and WSD / linear schedules
FareedKhan-dev committed 1 month ago
4b52229
feat(models): build either architecture from one config
FareedKhan-dev committed 1 month ago
ad9aaeb
0cca130
0f9596d
63b4299
70bc898
7554e51
170d4ac