sdreamforchen
sdreamforchen
for compactor_param, mask in compactor_mask_dict.items():####################单独设置compactor的梯度信息,加上lasso_grad梯度 compactor_param.grad.data = mask * compactor_param.grad.data lasso_grad = compactor_param.data * ((compactor_param.data ** 2).sum(dim=(1, 2, 3), keepdim=True) ** (-0.5))###########这个mask是乘以的loss第二项，和论文不同 compactor_param.grad.data.add_(resrep_config.lasso_strength, lasso_grad) if not if_accum_grad: if gradient_mask_tensor is...
我下游任务还是报TypeError: 'str' object cannot be interpreted as an integer 在服务器上运行，报警告： 2022-03-31 09:19:31.619078092 [W:onnxruntime:, graph.cc:3526 CleanUnusedInitializersAndNodeArgs] Removing initializer 'Constant_113'. It is not used by any node and should be removed from...
I'm confused about this below, in chapter 2.2： We tried two rescale strategies, first is to ensure the highest s2 is equal to the max IOU value among all possible...
class RepBN(nn.Layer): def __init__(self, channels): super(RepBN, self).__init__() self.alpha = self.create_parameter( shape=[1], attr=ParamAttr(initializer=Constant(value=1.)), dtype="float32") self.bn = nn.BatchNorm1D(channels) def forward(self, x): x = x.transpose([0, 2, 1]) x = self.bn(x) + self.alpha *...
您好，想请教一下，如何将训练的epoch/iteration配置到RT-DETR的head中呢？ 类似output=αA+（1-α）B，α=epoch/total_epoch；因为现有的architectures架构是将backbone、neck和head一起传到train里面的，不是很懂如何单独对head进行操作，且能传train中的epoch/iteration参数；能否给一个示例，谢谢！
Where is your loss，I don't find it
### 问题确认 Search before asking - [x] 我已经查询[历史issue](https://github.com/PaddlePaddle/PaddleDetection/issues)，没有类似需求。I have searched the [issues](https://github.com/PaddlePaddle/PaddleDetection/issues) and found no similar feature requests. ### 需求描述 Feature Description 不知道mamba在大尺寸图像方面有没有特殊的优势 ### 是否愿意提交PR Are you willing to submit...