Training error 'module' object has no attribute 'downsample'

[VirginiaMenezes]

When I am trying to run the train.py file for training, I am getting the error 'module' object has no attribute 'downsample'
I read that downsample in theano.tensor.signal import downsample has been moved to pool module.
But which function is equivalent in pool is equivalent to DownsampleFactorMax in downsample.

Please help ASAP.

Traceback (most recent call last):
  File "train.py", line 182, in <module>
    main(args)
  File "train.py", line 135, in main
    tfuncs, tvars = make_training_functions(cfg, model)
  File "train.py", line 31, in make_training_functions
    out = lasagne.layers.get_output(l_out, X)
  File "/usr/local/lib/python2.7/dist-packages/lasagne/layers/helper.py", line 185, in get_output
    all_outputs[layer] = layer.get_output_for(layer_inputs, **kwargs)
  File "/home/nvidia/virginia/voxnet/voxnet/layers.py", line 244, in get_output_for
    out = max_pool_3d(input, self.pool_shape)
  File "/home/nvidia/virginia/voxnet/voxnet/max_pool_3d.py", line 72, in max_pool_3d
    op = T.signal.downsample.DownsampleFactorMax((ds[1], ds[2]), ignore_border)
AttributeError: 'module' object has no attribute 'downsample'

[CATHUR]

I also encountered this situation, and I changed the source code as follows:
op = T.signal.pool.Pool((ds[1], ds[2]), ignore_border)

[stephencuinujia]

I got
ValueError: Pool mode parameter only support 'max', 'sum', 'average_inc_pad' and 'average_exc_pad'. Got False

After change to
op = T.signal.pool.Pool((ds[1], ds[2]), ignore_border)

Any idea?

[merium]

I solved the same issue by switching to theano 0.8.2

[buemi]

I solved this issue in theano 1.0.1 by replacing the following statements:

op = T.signal.downsample.DownsampleFactorMax((ds[1], ds[2]), ignore_border)
output = op(input_4D)

with these:

op = T.signal.pool.Pool(ignore_border)
output = op(input_4D, ws=(ds[1], ds[2]))

And the other call to DownsampleFactorMax in the same way.

[claell]

This is apparently caused by Theano moving downsample to a different package. I think Lasagne 0.1 uses the old location, which only works with Theano 0.8.2 or lower. Unfortunately there does not seem to be a more updated version of Lasagne.