'Module' object is not callable

[guoguoguilai]

I loaded the model from .json and .params， sym, arg_params, aux_params = mx.model.load_checkpoint(prefix[3], epoch[3]) mod = mx.mod.Module(symbol=L106_Net96_v2("test"), context=mx.gpu(), data_names=['data'], label_names=['landmark_target']) mod.bind(for_training=False, data_shapes=[('data', (batch_size[3], 3, 96, 96))]) mod.set_params(arg_params, aux_params) When I call pytorch_model = gluon2pytorch(mod, [(batch_size[3], 3, 96, 96)], dst_dir=None, pytorch_module_name='Test') Return: TypeError: 'Module' object is not callable

How can I solve this problem?How to convert module to be hybrid ?

[gmalivenko]

Hello @guoguoguilai. Can you provide me full source code so I can reproduce your error?

[guoguoguilai]

Hello @guoguoguilai. Can you provide me full source code so I can reproduce your error?

import mxnet as mx from gluon2pytorch import gluon2pytorch

def L106_Net96_v2(mode="train"): # def L106_Net96(mode="train"): """ Refine Network input shape 3 x 96 x 96 """ lnet106_basenum = 32 data = mx.symbol.Variable(name="data") landmark_target = mx.symbol.Variable(name="landmark_target")

conv1 = mx.symbol.Convolution(data=data, kernel=(3, 3), num_filter=lnet106_basenum, name="conv1")  # 96/94
bn1 = mx.sym.BatchNorm(data=conv1, name='bn1', fix_gamma=False, momentum=0.9)
prelu1 = mx.symbol.LeakyReLU(data=bn1, act_type="prelu", name="prelu1")

conv2_dw = mx.symbol.Convolution(data=prelu1, kernel=(2, 2), num_filter=lnet106_basenum, num_group=lnet106_basenum,
                                 name="conv2_dw")  # 94/93
bn2_dw = mx.sym.BatchNorm(data=conv2_dw, name='bn2_dw', fix_gamma=False, momentum=0.9)
prelu2_dw = mx.symbol.LeakyReLU(data=bn2_dw, act_type="prelu", name="prelu2_dw")
conv2_sep = mx.symbol.Convolution(data=prelu2_dw, kernel=(1, 1), num_filter=lnet106_basenum, name="conv2_sep")
bn2_sep = mx.sym.BatchNorm(data=conv2_sep, name='bn2_sep', fix_gamma=False, momentum=0.9)
prelu2_sep = mx.symbol.LeakyReLU(data=bn2_sep, act_type="prelu", name="prelu2_sep")

conv3_dw = mx.symbol.Convolution(data=prelu2_sep, kernel=(3, 3), stride=(2, 2), num_filter=lnet106_basenum,
                                 num_group=lnet106_basenum, name="conv3_dw")  # 93/46
bn3_dw = mx.sym.BatchNorm(data=conv3_dw, name='bn3_dw', fix_gamma=False, momentum=0.9)
prelu3_dw = mx.symbol.LeakyReLU(data=bn3_dw, act_type="prelu", name="prelu3_dw")
conv3_sep = mx.symbol.Convolution(data=prelu3_dw, kernel=(1, 1), num_filter=lnet106_basenum * 2, name="conv3_sep")
bn3_sep = mx.sym.BatchNorm(data=conv3_sep, name='bn3_sep', fix_gamma=False, momentum=0.9)
prelu3_sep = mx.symbol.LeakyReLU(data=bn3_sep, act_type="prelu", name="prelu3_sep")

conv4_dw = mx.symbol.Convolution(data=prelu3_sep, kernel=(2, 2), num_filter=lnet106_basenum * 2,
                                 num_group=lnet106_basenum * 2, name="conv4_dw")  # 46/45
bn4_dw = mx.sym.BatchNorm(data=conv4_dw, name='bn4_dw', fix_gamma=False, momentum=0.9)
prelu4_dw = mx.symbol.LeakyReLU(data=bn4_dw, act_type="prelu", name="prelu4_dw")
conv4_sep = mx.symbol.Convolution(data=prelu4_dw, kernel=(1, 1), num_filter=lnet106_basenum * 2, name="conv4_sep")
bn4_sep = mx.sym.BatchNorm(data=conv4_sep, name='bn4_sep', fix_gamma=False, momentum=0.9)
prelu4_sep = mx.symbol.LeakyReLU(data=bn4_sep, act_type="prelu", name="prelu4_sep")

conv5_dw = mx.symbol.Convolution(data=prelu4_sep, kernel=(3, 3), stride=(2, 2), num_filter=lnet106_basenum * 2,
                                 num_group=lnet106_basenum * 2, name="conv5_dw")  # 45/22
bn5_dw = mx.sym.BatchNorm(data=conv5_dw, name='bn5_dw', fix_gamma=False, momentum=0.9)
prelu5_dw = mx.symbol.LeakyReLU(data=bn5_dw, act_type="prelu", name="prelu5_dw")
conv5_sep = mx.symbol.Convolution(data=prelu5_dw, kernel=(1, 1), num_filter=lnet106_basenum * 2, name="conv5_sep")
bn5_sep = mx.sym.BatchNorm(data=conv5_sep, name='bn5_sep', fix_gamma=False, momentum=0.9)
prelu5_sep = mx.symbol.LeakyReLU(data=bn5_sep, act_type="prelu", name="prelu5_sep")

conv6_dw = mx.symbol.Convolution(data=prelu5_sep, kernel=(2, 2), num_filter=lnet106_basenum * 2,
                                 num_group=lnet106_basenum * 2, name="conv6_dw")  # 22/21
bn6_dw = mx.sym.BatchNorm(data=conv6_dw, name='bn6_dw', fix_gamma=False, momentum=0.9)
prelu6_dw = mx.symbol.LeakyReLU(data=bn6_dw, act_type="prelu", name="prelu6_dw")
conv6_sep = mx.symbol.Convolution(data=prelu6_dw, kernel=(1, 1), num_filter=lnet106_basenum * 2, name="conv6_sep")
bn6_sep = mx.sym.BatchNorm(data=conv6_sep, name='bn6_sep', fix_gamma=False, momentum=0.9)
prelu6_sep = mx.symbol.LeakyReLU(data=bn6_sep, act_type="prelu", name="prelu6_sep")

conv7_dw = mx.symbol.Convolution(data=prelu6_sep, kernel=(3, 3), stride=(2, 2), num_filter=lnet106_basenum * 2,
                                 num_group=lnet106_basenum * 2, name="conv7_dw")  # 21/10
bn7_dw = mx.sym.BatchNorm(data=conv7_dw, name='bn7_dw', fix_gamma=False, momentum=0.9)
prelu7_dw = mx.symbol.LeakyReLU(data=bn7_dw, act_type="prelu", name="prelu7_dw")
conv7_sep = mx.symbol.Convolution(data=prelu7_dw, kernel=(1, 1), num_filter=lnet106_basenum * 4, name="conv7_sep")
bn7_sep = mx.sym.BatchNorm(data=conv7_sep, name='bn7_sep', fix_gamma=False, momentum=0.9)
prelu7_sep = mx.symbol.LeakyReLU(data=bn7_sep, act_type="prelu", name="prelu7_sep")

conv8_dw = mx.symbol.Convolution(data=prelu7_sep, kernel=(2, 2), num_filter=lnet106_basenum * 4,
                                 num_group=lnet106_basenum * 4, name="conv8_dw")  # 10/9
bn8_dw = mx.sym.BatchNorm(data=conv8_dw, name='bn8_dw', fix_gamma=False, momentum=0.9)
prelu8_dw = mx.symbol.LeakyReLU(data=bn8_dw, act_type="prelu", name="prelu8_dw")
conv8_sep = mx.symbol.Convolution(data=prelu8_dw, kernel=(1, 1), num_filter=lnet106_basenum * 4, name="conv8_sep")
bn8_sep = mx.sym.BatchNorm(data=conv8_sep, name='bn8_sep', fix_gamma=False, momentum=0.9)
prelu8_sep = mx.symbol.LeakyReLU(data=bn8_sep, act_type="prelu", name="prelu8_sep")

conv9_dw = mx.symbol.Convolution(data=prelu8_sep, kernel=(3, 3), stride=(2, 2), num_filter=lnet106_basenum * 4,
                                 num_group=lnet106_basenum * 4, name="conv9_dw")  # 9/4
bn9_dw = mx.sym.BatchNorm(data=conv9_dw, name='bn9_dw', fix_gamma=False, momentum=0.9)
prelu9_dw = mx.symbol.LeakyReLU(data=bn9_dw, act_type="prelu", name="prelu9_dw")
conv9_sep = mx.symbol.Convolution(data=prelu9_dw, kernel=(1, 1), num_filter=lnet106_basenum * 8, name="conv9_sep")
bn9_sep = mx.sym.BatchNorm(data=conv9_sep, name='bn9_sep', fix_gamma=False, momentum=0.9)
prelu9_sep = mx.symbol.LeakyReLU(data=bn9_sep, act_type="prelu", name="prelu9_sep")

conv10_dw = mx.symbol.Convolution(data=prelu9_sep, kernel=(2, 2), num_filter=lnet106_basenum * 8,
                                  num_group=lnet106_basenum * 8, name="conv10_dw")  # 4/3
bn10_dw = mx.sym.BatchNorm(data=conv10_dw, name='bn10_dw', fix_gamma=False, momentum=0.9)
prelu10_dw = mx.symbol.LeakyReLU(data=bn10_dw, act_type="prelu", name="prelu10_dw")
conv10_sep = mx.symbol.Convolution(data=prelu10_dw, kernel=(1, 1), num_filter=lnet106_basenum * 8,
                                   name="conv10_sep")
bn10_sep = mx.sym.BatchNorm(data=conv10_sep, name='bn10_sep', fix_gamma=False, momentum=0.9)
prelu10_sep = mx.symbol.LeakyReLU(data=bn10_sep, act_type="prelu", name="prelu10_sep")

conv11_dw = mx.symbol.Convolution(data=prelu10_sep, kernel=(3, 3), num_filter=lnet106_basenum * 8,
                                  num_group=lnet106_basenum * 8, name="conv11_dw")  # 3/1
bn11_dw = mx.sym.BatchNorm(data=conv11_dw, name='bn11_dw', fix_gamma=False, momentum=0.9)
prelu11_dw = mx.symbol.LeakyReLU(data=bn11_dw, act_type="prelu", name="prelu11_dw")
conv11_sep = mx.symbol.Convolution(data=prelu11_dw, kernel=(1, 1), num_filter=lnet106_basenum * 8,
                                   name="conv11_sep")
bn11_sep = mx.sym.BatchNorm(data=conv11_sep, name='bn11_sep', fix_gamma=False, momentum=0.9)
prelu11_sep = mx.symbol.LeakyReLU(data=bn11_sep, act_type="prelu", name="prelu11_sep")

conv6_3 = mx.symbol.FullyConnected(data=prelu11_sep, num_hidden=212, name="conv6_3")
bn6_3 = mx.sym.BatchNorm(data=conv6_3, name='bn6_3', fix_gamma=False, momentum=0.9)
if mode == "test":
    landmark_pred = bn6_3
    group = mx.symbol.Group([landmark_pred])
else:

    landmark_pred = mx.symbol.LinearRegressionOutput(data=bn6_3, label=landmark_target,
                                                     grad_scale=1, name="landmark_pred")
    #landmark_pred = mx.symbol.
    out = mx.symbol.Custom(landmark_pred=landmark_pred, landmark_target=landmark_target,
                           op_type='negativemining_onlylandmark106', name="negative_mining")
    group = mx.symbol.Group([out])
#mx.viz.plot_network(landmark_pred, title='alexnet', save_format='jpg', hide_weights=True).view()
return group

batch_size = 4 prefix = "L106" epoch = 8000 sym, arg_params, aux_params = mx.model.load_checkpoint(prefix, epoch) mod = mx.mod.Module(symbol=L106_Net96_v2("test"), context=mx.gpu(), data_names=['data'], label_names=['landmark_target']) mod.bind(for_training=False, data_shapes=[('data', (batch_size, 3, 96, 96))]) mod.set_params(arg_params, aux_params) pytorch_model = gluon2pytorch(mod, [(batch_size, 3, 96, 96)], dst_dir=None, pytorch_module_name='Test')

Thank you Thank you in advance！

[ooguz166]

Same Error on this code too:

 from gluon2pytorch import gluon2pytorch

  if __name__ == '__main__':
      net = get_model(name="i3d_resnet50_v1_ucf101", nclass=101, pretrained=True)
      net.hybridize()
      net.collect_params().initialize()
      pytorch_model = gluon2pytorch(net, [(1, 3, 224, 224)], dst_dir=None)

TypeError: 'module' object is not callable

[ooguz166]

Same Error on this code too:

 from gluon2pytorch import gluon2pytorch

  if __name__ == '__main__':
      net = get_model(name="i3d_resnet50_v1_ucf101", nclass=101, pretrained=True)
      net.hybridize()
      net.collect_params().initialize()
      pytorch_model = gluon2pytorch(net, [(1, 3, 224, 224)], dst_dir=None)

TypeError: 'module' object is not callable

İf you build the lib from the setup.py importing like this "from gluon2pytorch.gluon2pytorch import gluon2pytorch" solves the issue