Edit时候超出显存

[hexi-ledi]

您好，我按照您的方法采用了--enable-model-cpu-offload，但是还是显示内存溢出，我的服务器是显存是24GB的，采用的命令格式是python scripts/inference.py --image scripts/people.png --instruction "Make her hair dark green."--flux-path /ICEdit-main/FLUX --lora-path /ICEdit-main/lora/pytorch_lora_weights.safetensors --enable-model-cpu-offload，请问有解决方案吗

[lxy2017]

我也遇到一样问题

[lxy2017]

用这个命令可以跑起来

Set environment variable to avoid memory fragmentation

export PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True

Run the inference script with the provided arguments

python scripts/inference.py --image assets/girl.png
--instruction "Make her hair dark green and her clothes checked."
--enable-model-cpu-offload

[hexi-ledi]

非常感觉您的回复，解决了部分问题

…

---原始邮件--- 发件人: ***@***.***> 发送时间: 2025年6月19日(周四) 中午1:55 收件人: ***@***.***>; 抄送: ***@***.******@***.***>; 主题: Re: [River-Zhang/ICEdit] Edit时候超出显存 (Issue #74) lxy2017 left a comment (River-Zhang/ICEdit#74) 用这个命令可以跑起来 Set environment variable to avoid memory fragmentation export PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True Run the inference script with the provided arguments python scripts/inference.py --image assets/girl.png --instruction "Make her hair dark green and her clothes checked." --enable-model-cpu-offload — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***>