Wenjie Qiang

Attention weight可视化

1

请问作者可以提供attention weight可视化的代码吗？从而方便后续工作分析网络在关注什么

Apply to other datasets

1

Excellent work ! P2PNet[1] is a SOTA method of crowd counting and effective method of crowd localization. Have you ever tried to train CPR + P2PNet on counting benchmarks (such...

how to load robust loss parameters

1

Hello, thank you very much for your elegant code. I want to use robust loss to train the model in the first stage and load the pre-trained model in the...

Warmup strategy

3

Hello, I would like to ask if the warmup strategy is not used, but the keep rate is directly set to the target value, will the experimental results differ greatly?

How to align with pytorch quantization .onnx models?

3

Hello, I did quantitative training using pytorch and exported the onnx model in the following way. I got a .onnx model. ` model=torchvision.models.quantization.resnet18(quantize=True) torch.export() ` In the meantime, I use...

多卡训练

3

请问可以多卡训练吗？我使用nn.DataParallel的话，loss就会特别大，mae也特别大。

请问可以多卡训练吗？我使用nn.DataParallel的话，loss就会特别大，mae也特别大。

关于网络的训练

4

作者您好，我理解deit_dge.py中，应该是使用了 Dynamic Grained Encoder。但是在代码dge.py中，sparse_train默认是false，并且这个参数在DynamicGrainedEncoder初始化的时候也没改，那么训练的时候走的是这个循环吗？ https://github.com/StevenGrove/vtpack/blob/38d5af0c3d9181177496619605ed1eb66c52445e/vtpack/layers/dge.py#L69 期待您的回复！