What should I do if I want to switch to another dataset, such as Duke and MSMT17?Does the teacher model need to be retrained?

[Be-yourself-zxy]

What should I do if I want to switch to another dataset, such as Duke and MSMT17?Does the teacher model need to be retrained?

[layumi]

Yes. You may use https://github.com/layumi/Person_reID_baseline_pytorch to train one teacher model.

[layumi]

I have provided the teacher setting at this issue #44

[Be-yourself-zxy]

Thank you for your reply,have a good day.------------------&nbsp;原始邮件&nbsp;------------------ 发件人:&nbsp;&quot;Zhedong&amp;nbsp;Zheng&quot;<[email protected]&gt; 发送时间:&nbsp;2020年7月16日(星期四) 晚上6:10 收件人:&nbsp;&quot;NVlabs/DG-Net&quot;<[email protected]&gt;; 抄送:&nbsp;&quot;Be-yourself-zxy&quot;<[email protected]&gt;;&quot;Author&quot;<[email protected]&gt;; 主题:&nbsp;Re: [NVlabs/DG-Net] What should I do if I want to switch to another dataset, such as Duke and MSMT17?Does the teacher model need to be retrained? (#46)

[Be-yourself-zxy]

I use https://github.com/layumi/Person_reID_baseline_pytorch to train one teacher model.
python train.py --name PCB --PCB --train_all --lr 0.02.
But the following error occurred：

Traceback (most recent call last):
File "train.py", line 48, in
trainer = DGNet_Trainer(config, gpu_ids)
File "/data/zxy/DG-Net-master/trainer.py", line 211, in init
teacher_model_tmp = load_network(model_tmp, teacher_name)
File "/data/zxy/DG-Net-master/trainer.py", line 159, in load_network
network.load_state_dict(torch.load(save_path))
File "/home/lei/.local/lib/python3.6/site-packages/torch/nn/modules/module.py", line 777, in load_state_dict
self.class.name, "\n\t".join(error_msgs)))
RuntimeError: Error(s) in loading state_dict for ft_net:
Missing key(s) in state_dict: "classifier.add_block.0.weight", "classifier.add_block.0.bias", "classifier.add_block.1.weight", "classifier.add_block.1.bias", "classifier.add_block.1.running_mean", "classifier.add_block.1.running_var", "classifier.classifier.0.weight", "classifier.classifier.0.bias".
Unexpected key(s) in state_dict: "classifier0.add_block.0.weight", "classifier0.add_block.0.bias", "classifier0.add_block.1.weight", "classifier0.add_block.1.bias", "classifier0.add_block.1.running_mean", "classifier0.add_block.1.running_var", "classifier0.add_block.1.num_batches_tracked", "classifier0.classifier.0.weight", "classifier0.classifier.0.bias", "classifier1.add_block.0.weight", "classifier1.add_block.0.bias", "classifier1.add_block.1.weight", "classifier1.add_block.1.bias", "classifier1.add_block.1.running_mean", "classifier1.add_block.1.running_var", "classifier1.add_block.1.num_batches_tracked", "classifier1.classifier.0.weight", "classifier1.classifier.0.bias", "classifier2.add_block.0.weight", "classifier2.add_block.0.bias", "classifier2.add_block.1.weight", "classifier2.add_block.1.bias", "classifier2.add_block.1.running_mean", "classifier2.add_block.1.running_var", "classifier2.add_block.1.num_batches_tracked", "classifier2.classifier.0.weight", "classifier2.classifier.0.bias", "classifier3.add_block.0.weight", "classifier3.add_block.0.bias", "classifier3.add_block.1.weight", "classifier3.add_block.1.bias", "classifier3.add_block.1.running_mean", "classifier3.add_block.1.running_var", "classifier3.add_block.1.num_batches_tracked", "classifier3.classifier.0.weight", "classifier3.classifier.0.bias", "classifier4.add_block.0.weight", "classifier4.add_block.0.bias", "classifier4.add_block.1.weight", "classifier4.add_block.1.bias", "classifier4.add_block.1.running_mean", "classifier4.add_block.1.running_var", "classifier4.add_block.1.num_batches_tracked", "classifier4.classifier.0.weight", "classifier4.classifier.0.bias", "classifier5.add_block.0.weight", "classifier5.add_block.0.bias", "classifier5.add_block.1.weight", "classifier5.add_block.1.bias", "classifier5.add_block.1.running_mean", "classifier5.add_block.1.running_var", "classifier5.add_block.1.num_batches_tracked", "classifier5.classifier.0.weight", "classifier5.classifier.0.bias".

[Be-yourself-zxy]

Can you share the trained models on the dataset MSMT17 and the dataset of MSMT17?

…

------------------&nbsp;原始邮件&nbsp;------------------ 发件人: "NVlabs/DG-Net" <[email protected]&gt;; 发送时间:&nbsp;2020年7月16日(星期四) 晚上6:13 收件人:&nbsp;"NVlabs/DG-Net"<[email protected]&gt;; 抄送:&nbsp;"我可以的"<[email protected]&gt;;"Author"<[email protected]&gt;; 主题:&nbsp;Re: [NVlabs/DG-Net] What should I do if I want to switch to another dataset, such as Duke and MSMT17?Does the teacher model need to be retrained? (#46) I have provided the teacher setting at this issue #44 — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.