pretrained model

[ai25395]

请问有预训练的模型提供吗，想快速测试一下效果。有的话发个链接可以吗，谢谢。

[ai25395]

训练非常慢，平台是9900k+2080ti，batchsize=1，大约22分钟1个epoch

[CVer-Yang]

我上传了模型在massachusetts道路数据集权重，链接：https://pan.baidu.com/s/13nboQ56F0SQuFSELyX7BQA  提取码：RCFS

…

---Original--- From: ***@***.***> Date: Mon, Apr 17, 2023 23:06 PM To: ***@***.***>; Cc: ***@***.***>; Subject: Re: [CVer-Yang/RCFSNet] pretrained model (Issue #4) 训练非常慢，平台是9900k+2080ti，batchsize=1，大约22分钟1个epoch — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[ai25395]

感谢回复。不过这个权重好像和项目里的RCFSNet不匹配，运行test时报错：
RuntimeError: Error(s) in loading state_dict for DataParallel:
E Missing key(s) in state_dict: "module.CDAM2.conv1.weight", .............."module.jiaohu1_Y.layer_norm.bias".

[CVer-Yang]

这个模型发表后为了便于理解修改了一些函数名字，可能导致模型权重不对应，我已上传一版原始模型文件（networks->RCFSNet-V1.py），可以实现权重对应，该文件内容较为冗余, 但主要部分与思想一致。

[axin1301]

Your work is really impressive. Can you please provide the pre-trained weights on the DeepGlobal dataset? Thank you.