`tx`

Usage:

$ tx [OPTIONS] COMMAND [ARGS]...

Options:

--install-completion: Install completion for the current shell.
--show-completion: Show completion for the current shell, to copy it or customize the installation.
--help: Show this message and exit.

Commands:

`tx train`

Train a model

Usage:

$ tx train [OPTIONS]

Options:

--model TEXT: HuggingFace model ID or local model path [required]
--dataset TEXT: HuggingFace dataset to use for training [required]
--loader TEXT: Loader used for loading the dataset [default: tx.loaders.text]
--split TEXT: The dataset split to use [default: train]
--output-dir PATH: The output directory where the model predictions and checkpoints will be written [required]
--load-checkpoint-path PATH: If specified, resume training from this checkpoint
--save-steps INTEGER: Number of steps between checkpoints [default: 500]
--max-steps INTEGER: The maximum number of training steps
--batch-size INTEGER: Batch size of each training batch [required]
--optimizer-args TEXT: Arguments for the optax optimizer [default: {"learning_rate": 1e-5, "weight_decay": 0.1}]
--tp-size INTEGER: Tensor parallelism degree to use for the model [default: 1]
--tracker [wandb]: Experiment tracker to report results to
--tracker-args TEXT: Arguments that will be passed to the experiment tracker (in JSON format) [default: {}]
--help: Show this message and exit.

Usage:

$ tx version [OPTIONS]

Options: