lora微调后合并模型出错
Author: GuoDragonCreated May 14, 2025Updated May 14, 2025
在进行lora微调合并模型时，使用
CUDA_VISIBLE_DEVICES=0 llamafactory-cli export
--model_name_or_path ......../chatglm3-6b
--adapter_name_or_path ....../lora/sft
--template chatglm3
--trust_remote_code
--finetuning_type lora
--export_dir ......../lora/merged_model
--export_size 1
--export_device cpu
--export_legacy_format true
报错ValueError: Dict is not supported.
Source: zai-org/ChatGLM3