FengWen

## pytorch 最小复现代码 ```python import torch.nn as nn def oneflow_cal(in_channels=512, out_channels=256, kernel_size=128, stride=1, groups=1, bias=False): padding = kernel_size // 2 conv = nn.Conv2d( in_channels, out_channels, kernel_size=kernel_size, stride=stride, padding=padding, groups=groups, bias=bias,...

> yolo不应该有128*128这么大的kernel吧，是不是有什么误会？ weight.shape=[256, 512, 128, 128]，光这一个conv就要8G显存。 找其他仓库对比了下，应该是代码的问题。

> 将 `os.environ["CUDA_VISIBLE_DEVICES"] = "1"` 放在`import oneflow`之前试试看? 放在`import oneflow`之前 可以生效

Here is a supported case that you can try using first https://github.com/siliconflow/onediff/pull/837 @joel-simon

> I meet a similar problem. Please try `cd onediff && git checkout dev_support_diffusers_ipa` @xiecon

> ## Code Your Diffusers version and please try @joel-simon ```python # pipe = compile_pipe(pipe) pipe.unet = oneflow_compile(pipe.unet) ```

get , Please try `pip3 install -U torch torchvision torchaudio` . torch > 2.0 and torch==1.13.1 different attention processors will be used ，Using Torch>2.0 version will solve this problem >...

> @ccssu 我有尝试切换到您提供的支持ip adapter 的分支，拉取代码后重新执行你在分支上的demo代码，还是会有这个问题AttributeError: 'list' object has no attribute 'to'，我diffuser是 0.27 torch 是最新的，cuda是12.1。谢谢 可以 设置下 `export ONEDIFF_DEBUG=1` 看下您那边 debug 模式下 日志么

> @ccssu ah gotcha, torch >= 2.0 did fix it, thanks! > > It's not working for multiple ip adapters but I don't think that's supported yet, right? Looking forward...

您好，onediff暂时不支持 python3.11， 请使用 python3.10 。 @cthulhu-tww > @ccssu 您好，打开debug日志后，出现了这样的报错 ERROR run got error: Check failed: (3 == 1) File "oneflow/core/job/job_interpreter.cpp", line