Mr Rotten peach

also trouble with this issue, is there any tips about how to train vit on ImageNet

The same question，maybe it is not a question，its a problem

I have no cuda, so we should I do if I want deploy the C++ api just for inference.

Maybe you need to check your feed data at 500 steps, I guess the size is not same as before so that the crashes occur.

> @RaceSu I am having the same error. Did you find the solution? I solved it, the image height and width you set in code is not same as the...

Maybe you need to modify here, mobilenetv3_seg.py script *around 60 line* class _LRASPP(nn.Module): """Lite R-ASPP""" def __init__(self, in_channels, norm_layer, **kwargs): super(_LRASPP, self).__init__() out_channels = 128 self.b0 = nn.Sequential( nn.Conv2d(in_channels, out_channels,...

> When i exported onnx model, onnx cannot surpport adaptive_avg_pool2d.How can i fix the problem? upgrade your onnx version, use opset=13 or fix your input size use nn.AvgPool2d with suitable...

same issue

yolov3是有的，yolov4也很容易部署

> 1.假设我从摄像头获取数据，VPSS可以直接输出resize后的YUV数据，理论可以省掉opencv的数据处理层。 > 2.输入端为8bit 正整数，一般的模型均为float32，这样的条件下，如果做量化效果会比较好 参考下文章： https://openaccess.thecvf.com/content_cvpr_2018/papers/Jacob_Quantization_and_Training_CVPR_2018_paper.pdf