returnn-common-generated-config.py

from returnn.tf.util.data import Dim, batch_dim, single_step_dim, SpatialDim, FeatureDim

use_tensorflow = True
behavior_version = 12

extern_data_data_dim_tags_1_time_dim = SpatialDim('time')
extern_data_data_dim_tags_2_input_dim = FeatureDim('input', 13)
network_linear_subnetwork_random_shape_1_linear_out_dim = FeatureDim('linear-out', 13)

extern_data = {
  'data': {
    'dim_tags': (
      batch_dim,
      extern_data_data_dim_tags_1_time_dim,
      extern_data_data_dim_tags_2_input_dim
    ),
    'dtype': 'float32',
    'available_for_inference': True
  }
}

network = {
  'linear': {
    'class': 'subnetwork',
    'from': [],
    'subnetwork': {
      'random': {
        'class': 'random',
        'shape': (
          extern_data_data_dim_tags_2_input_dim,
          network_linear_subnetwork_random_shape_1_linear_out_dim
        ),
        'distribution': 'uniform',
        'minval': -0.4803844614152614,
        'maxval': 0.4803844614152614,
        'dtype': 'float32',
        'static': True,
        'out_shape': {extern_data_data_dim_tags_2_input_dim, network_linear_subnetwork_random_shape_1_linear_out_dim}
      },
      'dot': {
        'class': 'dot',
        'from': ['base:data:data', 'weight'],
        'reduce': extern_data_data_dim_tags_2_input_dim,
        'out_shape': {batch_dim, extern_data_data_dim_tags_1_time_dim, network_linear_subnetwork_random_shape_1_linear_out_dim}
      },
      'add': {
        'class': 'combine',
        'from': ['dot', 'bias'],
        'kind': 'add',
        'out_shape': {batch_dim, extern_data_data_dim_tags_1_time_dim, network_linear_subnetwork_random_shape_1_linear_out_dim}
      },
      'output': {
        'class': 'copy',
        'from': 'add',
        'out_shape': {batch_dim, extern_data_data_dim_tags_1_time_dim, network_linear_subnetwork_random_shape_1_linear_out_dim}
      },
      'weight': {
        'class': 'variable',
        'shape': [
          extern_data_data_dim_tags_2_input_dim,
          network_linear_subnetwork_random_shape_1_linear_out_dim
        ],
        'param_name': 'param',
        'init_by_layer': 'random'
      },
      'bias': {
        'class': 'variable',
        'shape': [
          network_linear_subnetwork_random_shape_1_linear_out_dim
        ],
        'param_name': 'param',
        'init': 0.0
      }
    },
    'out_shape': {batch_dim, extern_data_data_dim_tags_1_time_dim, network_linear_subnetwork_random_shape_1_linear_out_dim}
  },
  'output': {
    'class': 'copy',
    'from': 'linear',
    'out_shape': {batch_dim, extern_data_data_dim_tags_1_time_dim, network_linear_subnetwork_random_shape_1_linear_out_dim}
  }
}