hq_wei
hq_wei
I have trained it however I get the following error: RuntimeError: element 0 of tensors does not require grad and does not have a grad_fn 
Is there any one having tried training?
词向量文件链接: https://pan.baidu.com/s/1NgaZrA-XyA7HKHDdowHFDw 提取码: n5ya
> 请问lstm训练中idsmatrix就是25000条训练数据对应wordvecor的id吗。 data文件夹中的数据在训练过程中没有使用对么。 对，直接加载的转换好的词向量ｎｐｙ文件：ids = np.load('./training_data/idsMatrix.npy')
@EthanZhangYi Have you reproduced the source-only result? I use VGG with an image size of 1280x720. When testing in the target domain, the image size is resize to 1024x512. But...
@EthanZhangYi Have you reproduced the source-only based on resnet101(PSPnet, deeplabV2 or deeplabV3)? I have trained on Cityscapes and valuated on GTA5, I get 30.28(MIOU). However, when I trained on GTA5...
I also encountered the same problem
我试过比较有效的RCNN 部分分类和回归的蒸馏（和FPN没什么关系），在backbone部分做特征的mimic时，需要加个卷积层进行变换到相同大小。
我这是也是一个卷积层进行通道和featuremap大小变化，大小一样就可以计算相似度了： `import torch.nn as nn import torch.nn.functional as F class Stu_Feature_Adap(nn.Module): def __init__(self,input_channel=256, output_channel=1024,kernel_size=2,padding=0): super(Stu_Feature_Adap, self).__init__() self.conv1 = nn.Conv2d(input_channel, output_channel, kernel_size=kernel_size, padding=padding) self.relu = nn.ReLU() def forward(self, x): x =...
swift.llm.template.base.MaxLengthError: Current length of row(103074) is larger than the max_length(40960).我也遇到了，Qwen3-4B没有，8B模型就有这个问题，说明是输出长度超长？还没截断？