Ziqing Yang comments

Results 212 comments of


                                            Ziqing Yang

deepspeed训练报错

> @airaria 重新生成同样的报错参考iMountTai的建议 > 关注一下内存的变化，可能是内存不足

运行报错：RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cuda:1! (when checking argument for argument index in method wrapper__index_select)

提供一下启动参数？

运行报错：RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cuda:1! (when checking argument for argument index in method wrapper__index_select)

建议使用deepspeed，可以参考我们最新的[预训练提交脚本](https://github.com/ymcui/Chinese-LLaMA-Alpaca/wiki/预训练脚本)

Can't find adapter_config.json

[参考wiki](https://github.com/ymcui/Chinese-LLaMA-Alpaca/wiki/指令精调脚本#训练后文件整理)

请问通过run_pt.sh后得到的参数，哪个文件是参数的增量呀

新建一个文件夹，把`pytorch_model.bin`放进去并改名`adapter_model.bin`，并补齐tokenizer相关和config相关文件，使得文件夹内容与我们发布的如Chinese-LLaMA-LoRA-7b一致。大致流程如下： ```bash mkdir lora_model cp pytorch_model.bin lora_model/adapter_model.bin cp Chinese-LLaMA-LoRA-7b/adapter_config.json lora_model/ cp Chinese-LLaMA-LoRA-7b/*token* lora_model/ ``` 其中你需要修改`adapter_config.json`中的LoRA参数，以和你训练时用的参数保持一致。之后就可以用merge_llama_with_chinese_lora.py合并了我们之后会在wiki中更新相关流程说明。

Ziqing Yang

deepspeed训练报错

运行报错：RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cuda:1! (when checking argument for argument index in method wrapper__index_select)

运行报错：RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cuda:1! (when checking argument for argument index in method wrapper__index_select)

Can't find adapter_config.json

请问通过run_pt.sh后得到的参数，哪个文件是参数的增量呀

请问通过run_pt.sh后得到的参数，哪个文件是参数的增量呀

回答的问题一直重复

合并lora模型和原始模型出问题

如何用transformer加载量化后的模型

Wiki: 与LangChain进行集成文档错误