ValueError: Expected input batch_size (98304) to match target batch_size (24576).

[ssbilakeri]

when I use MDRS2_config file as overwrite_file getting bellow error. please suggest me where Im wrong

Traceback (most recent call last):
File "demo.py", line 11, in
trainer.train()
File "..\MDRSREID\Trainer\MDRSReIDTrainer.py", line 111, in train
source_loss += self.loss_functions[loss_type.name](item, pred, step=self.current_step)['loss']
File "..\MDRSREID\Loss_Meter\Multi_Seg_loss.py", line 24, in call
loss_list = [self.criterion(multi_seg_pred_list[i].permute(0, 2, 3, 1).contiguous().view(-1, C), ps_label) for i in range(len(multi_seg_pred_list))]
File "..\MDRSREID\Loss_Meter\Multi_Seg_loss.py", line 24, in
loss_list = [self.criterion(multi_seg_pred_list[i].permute(0, 2, 3, 1).contiguous().view(-1, C), ps_label) for i in range(len(multi_seg_pred_list))]
File "C:\Users\ssbilakeri\Anaconda3\lib\site-packages\torch\nn\modules\module.py", line 489, in call
result = self.forward(*input, **kwargs)
File "C:\Users\ssbilakeri\Anaconda3\lib\site-packages\torch\nn\modules\loss.py", line 904, in forward
ignore_index=self.ignore_index, reduction=self.reduction)
File "C:\Users\ssbilakeri\Anaconda3\lib\site-packages\torch\nn\functional.py", line 1970, in cross_entropy
return nll_loss(log_softmax(input, 1), target, weight, None, ignore_index, None, reduction)
File "C:\Users\ssbilakeri\Anaconda3\lib\site-packages\torch\nn\functional.py", line 1788, in nll_loss
.format(input.size(0), target.size(0)))
ValueError: Expected input batch_size (98304) to match target batch_size (24576).

[nickhuang1996]

@ssbilakeri
For this bugs, I have rewritten the code for overwrite_configs and modified the MDRS_seg_loss.py for two options, which is correponding to backbone options.

[qianqiansbaby]

@ssbilakeri
Hi, I have same problems but I choose use option 2 then no errors and I get the same results better than those in this paper. Could you please tell me how to choose these overwrite config files for training please!:relaxed:

[qianqiansbaby]

@nickhuang1996
I have another question：
How the RamdomIdentitySampler2.py works?
I can't understand how to use this sampler.:sob: Please tell me how to use it for training.

[nickhuang1996]

@qianqiansbaby
This RamdomIdentitySampler is to select identities and their images. For example, if you set identities to 4 and their images to 8, then the batch size for training is 4*8 = 32.

[ssbilakeri]

Resolved error. Thank you.

…

On Thu, 15 Oct 2020, 7:10 pm nickhuang1996, ***@***.***> wrote: Assigned #2 <#2> to @ssbilakeri <https://github.com/ssbilakeri>. — You are receiving this because you were assigned. Reply to this email directly, view it on GitHub <#2 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AOOYJBPKUTEKKX57QZ3MAX3SK33UXANCNFSM4SQ5XRQQ> .

[nickhuang1996]

@ssbilakeri
Good luck!:relaxed: You can use my MDRS_feat_dict.mat files to test and no problem. Thank you for cooperating with me for modification on this code.:stuck_out_tongue_winking_eye:

[D77-d77]

@nickhuang1996
Thank you for resolving errors.😘 I have achieved excellent results through your feat_dict files.👍