Check failed: new_queue_pairs_.size() == 0

[paraggoel14]

Hi,

I am running the ImageNet example from the website, and while running caffe, I run into this error after it reads in the mean file:

data_reader.cpp:98] Check failed: new_queue_pairs_.size() == 0 (1 vs. 0)

I am not sure what to make of this error. What can be the possible reason behind this?

[seanbell]

The documentation for the error might answer your question:

caffe/src/caffe/data_reader.cpp

Lines 94 to 98 in 196b995

	// Check no additional readers have been created. This can happen if
	// more than one net is trained at a time per process, whether single
	// or multi solver. It might also happen if two data layers have same
	// name and same source.
	CHECK_EQ(new_queue_pairs_.size(), 0);

[etienne87]

Hi, i have the same error... my configuration is like this :

layer {
name: "image0" type: "Data"
top: "data"
data_param {
source: "dataXXX/image.lmdb"
batch_size: 32
backend: LMDB
}
}

layer {
name: "label0" type: "Data"
top: "label"
data_param {
source: "dataXXX/label.lmdb"
batch_size: 32
backend: LMDB
}
}

... rest is definition of the convnet...

What is the problem? should I specify include{phase: TRAIN} somewhere?

[ultrai]

Same error with similar data blob for image segmentation

[karunaahuja]

Same error. Any quick fix? It is very urgent.

[lukeyeager]

If your data layers look like @etienne87's above, then you need to mark one as your TRAIN data and another as your TEST data. Look at the reference networks for guidance:

https://github.com/BVLC/caffe/blob/rc3/models/bvlc_alexnet/train_val.prototxt#L7-L28
https://github.com/BVLC/caffe/blob/rc3/models/bvlc_googlenet/train_val.prototxt#L7-L30
https://github.com/BVLC/caffe/blob/rc3/examples/mnist/lenet_train_test.prototxt#L7-L26

layer {
  name: "training_data"
  type: "Data"
  top: "data"
  top: "label"
  include {
    phase: TRAIN
  }
  # ...
}
layer {
  name: "validation_data"
  type: "Data"
  top: "data"
  top: "label"
  include {
    phase: TEST
  }
  # ...
}