Ssd.pytorch: got this Error while training on standard data

[Kfirbarzilay]

got this Error while training in the 510 iteration:
""" File "train.py", line 255, in
train()
File "train.py", line 165, in train
images, targets = next(batch_iterator)
File "/usr/local/lib/python3.5/dist-packages/torch/utils/data/dataloader.py", line 271, in __next__
raise StopIteration
StopIteration
"""
what may be causing it?

[whang94]

I met the same problem

[ghost]

I met the same problem

[rbackman22]

I have the same problem.
Please note: I am a bit of a novice at this - I have never tried to "fix" anything before...
Also had one other issue, had to copy ssd.pytorch-master/data/coco/coco_labels.txt to data/coco/
(who knows what that did)

I tried this, and it seems to work (maybe):
I altered line in file train.py:
images, targets = next(batch_iterator)
to:
try:
images, targets = next(batch_iterator)
except StopIteration:
print('Batch complete')
break

To my surprise it created weights/VOC.pth file!!!
Then I merrily ran: python3 eval.py --trained_model weights/VOC.pth
It takes a while to plough through 4952 images...

I got: Mean AP = 0.0112
I think that is too low - hence I don't think I created a "fix"...
So I downloaded weights/ssd300_mAP_77.43_v2.pth
And ran: python3 eval.py
It runs quite a bit faster...
It got: Mean AP = 0.7749
Conclusion: My "fix" did not work, but it was fun while it lasted...

Now we wait for someone that knows something to help us...

[rbackman22]

Iteration is supposed to go to 120000.
My "fix" was stopping after 510 iterations

Hence I have now changed line in train.py from:
    images, targets = next(batch_iterator)
to:
    try:
        images, targets = next(batch_iterator)
    except StopIteration:
        batch_iterator = iter(data_loader)
        images, targets = next(batch_iterator)

Will see if the evaluation improves...

[ghost]

The problem is not to iterate data_loader.
For the time being, you could fix it with

  batch_iterator = None
  for iteration in range(args.start_iter, cfg['max_iter']):
        if (not batch_iterator) or (iteration % epoch_size ==0):
            batch_iterator = iter(data_loader)
            loc_loss = 0
            conf_loss = 0
            epoch += 1

[Usernamezhx]

@rbackman22 YEAH!! it works for me .I am waiting for the result.

[funnyuday]

it's work for me too.Thank you very much!

[yhcao6]

This also solve my problem. You are awesome man!

[shenyann]

@rbackman22 @Usernamezhx Hey guys, how is it going? Did you guys get better results and faster inference speed after change the try expect part?

"""
I got: Mean AP = 0.0112
I think that is too low - hence I don't think I created a "fix"...
"""

[YongJang]

@rbackman22 It works! thank you

[niaoyu]

The problem is not to iterate data_loader.
For the time being, you could fix it with

  batch_iterator = None
  for iteration in range(args.start_iter, cfg['max_iter']):
        if (not batch_iterator) or (iteration % epoch_size ==0):
            batch_iterator = iter(data_loader)
            loc_loss = 0
            conf_loss = 0
            epoch += 1

This should be the correct answer

[tanlizhi]

I have the same problem.
Please note: I am a bit of a novice at this - I have never tried to "fix" anything before...
Also had one other issue, had to copy ssd.pytorch-master/data/coco/coco_labels.txt to data/coco/
(who knows what that did)

I tried this, and it seems to work (maybe):
I altered line in file train.py:
images, targets = next(batch_iterator)
to:
try:
images, targets = next(batch_iterator)
except StopIteration:
print('Batch complete')
break

To my surprise it created weights/VOC.pth file!!!
Then I merrily ran: python3 eval.py --trained_model weights/VOC.pth
It takes a while to plough through 4952 images...

I got: Mean AP = 0.0112
I think that is too low - hence I don't think I created a "fix"...
So I downloaded weights/ssd300_mAP_77.43_v2.pth
And ran: python3 eval.py
It runs quite a bit faster...
It got: Mean AP = 0.7749
Conclusion: My "fix" did not work, but it was fun while it lasted...

Now we wait for someone that knows something to help us...

HEY,did you work out it? I got the same problem and cannot found some wrong with it?

[douliyoutang0529]

@small-mayfly HEY,have you solved it?