cli.main

cli.main

Click CLI definitions for various axolotl commands.

Functions

Name	Description
agent_docs	Show agent-optimized documentation.
cli	Axolotl CLI - Train and fine-tune large language models
config_schema	Dump the full config JSON schema.
evaluate	Evaluate a model.
fetch	Fetch example configs or other resources.
inference	Run inference with a trained model.
merge_lora	Merge trained LoRA adapters into a base model.
merge_sharded_fsdp_weights	Merge sharded FSDP model weights.
preprocess	Preprocess datasets before training.
train	Train or fine-tune a model.

agent_docs

cli.main.agent_docs(topic, list_topics)

Show agent-optimized documentation.

Prints reference docs designed for AI coding agents. These docs are bundled with the package — no network access needed.

 Examples: axolotl agent-docs # overview (start here) axolotl agent-docs grpo # GRPO reference axolotl agent-docs sft # SFT reference axolotl agent-docs –list # list all topics

cli

cli.main.cli()

Axolotl CLI - Train and fine-tune large language models

config_schema

cli.main.config_schema(output_format, field)

Dump the full config JSON schema.

Useful for AI agents and tooling to discover all available config options, their types, defaults, and descriptions.

 Examples: axolotl config-schema # full JSON schema axolotl config-schema –format yaml # YAML format axolotl config-schema –field adapter # single field

evaluate

cli.main.evaluate(ctx, config, launcher, **kwargs)

Evaluate a model.

Parameters

Name	Type	Description	Default
ctx	click.Context	Click context for extra args.	required
config	str	Path to axolotl config YAML file.	required
launcher	str	Launcher to use for multi-GPU evaluation (“accelerate”, “torchrun”, or “python”).	required
kwargs		Additional keyword arguments which correspond to CLI args or axolotl config options.	{}

fetch

cli.main.fetch(directory, dest)

Fetch example configs or other resources.

Available directories: - examples: Example configuration files - deepspeed_configs: DeepSpeed configuration files - docs: Full documentation (Quarto markdown files)

Parameters

Name	Type	Description	Default
directory	str	One of examples, deepspeed_configs, docs.	required
dest	Optional[str]	Optional destination directory.	required

inference

cli.main.inference(ctx, config, launcher, gradio, **kwargs)

Run inference with a trained model.

Parameters

Name	Type	Description	Default
ctx	click.Context	Click context for extra args.	required
config	str	Path to axolotl config YAML file.	required
launcher	str	Launcher to use for multi-GPU inference (“accelerate”, “torchrun”, or “python”).	required
gradio	bool	Whether to use Gradio browser interface or command line for inference.	required
kwargs		Additional keyword arguments which correspond to CLI args or axolotl config options.	{}

merge_lora

cli.main.merge_lora(config, **kwargs)

Merge trained LoRA adapters into a base model.

Parameters

Name	Type	Description	Default
config	str	Path to axolotl config YAML file.	required
kwargs		Additional keyword arguments which correspond to CLI args or axolotl config options.	{}

merge_sharded_fsdp_weights

cli.main.merge_sharded_fsdp_weights(ctx, config, launcher, **kwargs)

Merge sharded FSDP model weights.

Parameters

Name	Type	Description	Default
ctx	click.Context	Click context for extra args.	required
config	str	Path to axolotl config YAML file.	required
launcher	str	Launcher to use for weight merging (“accelerate”, “torchrun”, or “python”).	required
kwargs		Additional keyword arguments which correspond to CLI args or axolotl config options.	{}

preprocess

cli.main.preprocess(config, cloud=None, **kwargs)

Preprocess datasets before training.

Parameters

Name	Type	Description	Default
config	str	Path to axolotl config YAML file.	required
cloud	Optional[str]	Path to a cloud accelerator configuration file.	None
kwargs		Additional keyword arguments which correspond to CLI args or axolotl config options.	{}

train

cli.main.train(
    ctx,
    config,
    launcher='accelerate',
    cloud=None,
    sweep=None,
    **kwargs,
)

Train or fine-tune a model.

Parameters

Name	Type	Description	Default
ctx	click.Context	Click context for extra args.	required
config	str	Path to axolotl config YAML file.	required
launcher	Literal['accelerate', 'torchrun', 'python']	Launcher to use for multi-GPU training (“accelerate”, “torchrun”, or “python”).	'accelerate'
cloud	str | None	Path to a cloud accelerator configuration file	None
sweep	str | None	Path to YAML config for sweeping hyperparameters.	None
kwargs		Additional keyword arguments which correspond to CLI args or axolotl config options.	{}