This is because of what

[tom666tom666]

Traceback (most recent call last):
File "train.py", line 275, in
train()
File "train.py", line 198, in train
loss_l, loss_c = criterion(out, targets)
File "/usr/local/lib/python2.7/dist-packages/torch/nn/modules/module.py", line 491, in call
result = self.forward(*input, **kwargs)
File "/home/qiuyj/SSD/layers/modules/multibox_loss.py", line 101, in forward
loss_c[pos] = 0 # filter out pos boxes for now
RuntimeError: The shape of the mask [1, 8732] at index 0 does not match the shape of the indexed tensor [8732, 1] at index 0

use VOC dataset
`

[qigtang]

having the same problem here.

[wjourney]

I have the same problem. Have you solved it?

[qigtang]

I changed the code the following way, seems working fine for me.

@@ -95,8 +95,8 @@ class MultiBoxLoss(nn.Module):
         loss_c = log_sum_exp(batch_conf) - batch_conf.gather(1, conf_t.view(-1, 1))

         # Hard Negative Mining
-        loss_c[pos] = 0  # filter out pos boxes for now
         loss_c = loss_c.view(num, -1)
+        loss_c[pos] = 0  # filter out pos boxes for now
         _, loss_idx = loss_c.sort(1, descending=True)
         _, idx_rank = loss_idx.sort(1)
         num_pos = pos.long().sum(1, keepdim=True)

[yuanshengyu]

I changed the code the following way, seems working fine for me.

@@ -95,8 +95,8 @@ class MultiBoxLoss(nn.Module):
         loss_c = log_sum_exp(batch_conf) - batch_conf.gather(1, conf_t.view(-1, 1))

         # Hard Negative Mining
-        loss_c[pos] = 0  # filter out pos boxes for now
         loss_c = loss_c.view(num, -1)
+        loss_c[pos] = 0  # filter out pos boxes for now
         _, loss_idx = loss_c.sort(1, descending=True)
         _, idx_rank = loss_idx.sort(1)
         num_pos = pos.long().sum(1, keepdim=True)

it works fine for me too. But I meet another problem, Loss keeps NAN

[J0hannB]

it works fine for me too. But I meet another problem, Loss keeps NAN

Same problem here.

Maybe we're doing something else wrong that causes the error in the first place? Moving the loss_c[pos] = 0 line might be fixing a symptom, not the original issue.

[J0hannB]

I was able to fix my nan issue.

My solution: #173 (comment)