Arguments#

class FlagEmbedding.finetune.embedder.decoder_only.icl.DecoderOnlyEmbedderICLModelArguments(model_name_or_path: str, config_name: str | None = None, tokenizer_name: str | None = None, cache_dir: str | None = None, trust_remote_code: bool = False, token: str = <factory>, peft_model_path: str = '', use_lora: bool = True, lora_rank: int = 64, lora_alpha: float = 16, lora_dropout: float = 0.1, target_modules: ~typing.List[str] = <factory>, use_flash_attn: bool = False, use_slow_tokenizer: bool = False, from_peft: str | None = None, modules_to_save: str | None = None, raw_peft: str | None = None, additional_special_tokens: str | None = None, save_merged_lora_model: bool = False)[source]#

Model argument class for decoder only icl model.

class FlagEmbedding.finetune.embedder.decoder_only.icl.DecoderOnlyEmbedderICLDataArguments(train_data: str | None = None, cache_path: str | None = None, train_group_size: int = 8, query_max_len: int = 32, passage_max_len: int = 128, pad_to_multiple_of: int | None = None, max_example_num_per_dataset: int = 100000000, query_instruction_for_retrieval: str | None = None, query_instruction_format: str = '{}{}', knowledge_distillation: bool = False, passage_instruction_for_retrieval: str | None = None, passage_instruction_format: str | None = '{}{}', shuffle_ratio: float = 0.0, same_dataset_within_batch: bool = False, small_threshold: int = 0, drop_threshold: int = 0, example_query_max_len: int = 64, example_passage_max_len: int = 96, retrieval_use_examples: bool = True, icl_suffix_str: str = '\nResponse:')[source]#

Data argument class for decoder only icl model.