| import logging | |
| def validate_config(cfg): | |
| if cfg.load_4bit: | |
| raise ValueError("cfg.load_4bit parameter has been deprecated and replaced by cfg.gptq") | |
| if cfg.adapter == "qlora": | |
| if cfg.merge_lora: | |
| # can't merge qlora if loaded in 8bit or 4bit | |
| assert cfg.load_in_8bit is not True | |
| assert cfg.gptq is not True | |
| assert cfg.load_in_4bit is not True | |
| else: | |
| assert cfg.load_in_8bit is not True | |
| assert cfg.gptq is not True | |
| assert cfg.load_in_4bit is True | |
| if not cfg.load_in_8bit and cfg.adapter == "lora": | |
| logging.warning("We recommend setting `load_in_8bit: true` for LORA finetuning") | |
| # TODO | |
| # MPT 7b | |
| # https://github.com/facebookresearch/bitsandbytes/issues/25 | |
| # no 8bit adamw w bf16 | |