cli.utils.load

cli.utils.load

Utilities for model, tokenizer, etc. loading.

Functions

Name	Description
load_model_and_tokenizer	Helper function for loading a model, tokenizer, and processor specified in the
resolve_chat_template_str	Resolves the chat template string for inference from the axolotl config,

load_model_and_tokenizer

cli.utils.load.load_model_and_tokenizer(cfg, inference=False)

Helper function for loading a model, tokenizer, and processor specified in the given axolotl config.

Parameters

Name	Type	Description	Default
cfg	DictDefault	Dictionary mapping axolotl config keys to values.	required
inference	bool	Boolean denoting inference mode.	False

Returns

Name	Type	Description
	tuple[PreTrainedModel, PreTrainedTokenizer | PreTrainedTokenizerFast | Any, ProcessorMixin | None]	Tuple of (PreTrainedModel, PreTrainedTokenizer, ProcessorMixin).

resolve_chat_template_str

cli.utils.load.resolve_chat_template_str(cfg, tokenizer)

Resolves the chat template string for inference from the axolotl config, mirroring how it would be resolved at training time: an explicit chat_template config takes precedence, then the first dataset’s chat_template if that dataset is of type chat_template.

Parameters

Name	Type	Description	Default
cfg	DictDefault	Dictionary mapping axolotl config keys to values.	required
tokenizer	PreTrainedTokenizer | PreTrainedTokenizerFast | Any	Tokenizer to fall back to for tokenizer-default templates.	required

Returns

Name	Type	Description
	str | None	Chat template string, or None if the config does not specify one.