Train your own language model with one size dial.
Trains a language model with nanochat and guides evaluation and chat with the resulting model.
When to use it
Use when setting up nanochat, training or finetuning a model, configuring its depth, or chatting with it.
Give it a model-training request; it sets up nanochat and runs a pipeline that produces a trained model.
Uses uv for dependency management; installation instructions are included if it is missing.
Requires Python to run the training entry point; no version is specified.
This skill
wandb
Logs training metrics to wandb
uv
Installs uv from a remote script