1518630367

> 显存不够导致模型中有些参数是 meta tensor，在命令后加一下 `--device cpu` 按道理来说A100 80G 不应该会显存不够，看了一下显存没有占到1/8的时候就报错了，然后如果使用cpu会出现以下报错 pth_to_hf.py: error: unrecognized arguments: --device cpu

> 显存不够导致模型中有些参数是 meta tensor，在命令后加一下 `--device cpu` 使用旧版的xtuner 可以秒合并，新版的就会出现那个问题

import torch from datasets import load_dataset from mmengine.config import read_base from mmengine.dataset import DefaultSampler from mmengine.hooks import (CheckpointHook, DistSamplerSeedHook, IterTimerHook, LoggerHook, ParamSchedulerHook) from mmengine.optim import AmpOptimWrapper, CosineAnnealingLR from peft import...

> Can pdfminer convert xml to pdf? Because I need to use pdfminer to convert pdf to xml, and then make some modifications before converting it back to pdf The...