[BUG] <使用ModelScope进行推理报错：非法指令 (核心已转储)>

hwddzx commented 7 months ago

是否已有关于该错误的issue或讨论？ | Is there an existing issue / discussion for this?

[X] 我已经搜索过已有的issues和讨论 | I have searched the existing issues / discussions

该问题是否在FAQ中有解答？ | Is there an existing answer for this in FAQ?

[X] 我已经搜索过FAQ | I have searched FAQ

当前行为 | Current Behavior

在loading完成之后就会出现错误，并且会导致我的python虚拟环境不可用，web UI之前是正常使用，之后也不可用。

Loading checkpoint shards: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 10/10 [00:20<00:00, 2.06s/it] We've detected an older driver with an RTX 4000 series GPU. These drivers have issues with P2P. This can affect the multi-gpu inference when using accelerate device_map.Please make sure to update your driver to the latest version which resolves this. The model is automatically converting to bf16 for faster inference. If you want to disable the automatic precision, please manually add bf16/fp16/fp32=True to "AutoModelForCausalLM.from_pretrained". Loading checkpoint shards: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 10/10 [00:19<00:00, 1.92s/it] We've detected an older driver with an RTX 4000 series GPU. These drivers have issues with P2P. This can affect the multi-gpu inference when using accelerate device_map.Please make sure to update your driver to the latest version which resolves this. 非法指令 (核心已转储)

期望行为 | Expected Behavior

期望能正常推理

复现方法 | Steps To Reproduce

安装PyTorch==2.2.2
pip install requirements.txt
执行ModelScope的对话demo

运行环境 | Environment

- OS: Ubuntu 22
- Python:10
- Transformers:
- PyTorch:2.2.2
- CUDA (`python -c 'import torch; print(torch.version.cuda)'`):12.4

备注 | Anything else?

No response

frostyplanet commented 6 months ago

We've detected an older driver with an RTX 4000 series GPU 这行warning，在我升级 nvidia-driver-550 之后消失了@hwddzx 你这边是什么显卡配置, 什么驱动版本？具体模型和参数是哪个？

xmaple04 commented 3 months ago

我也遇到这问题了，两张4090显卡

QwenLM / Qwen-VL