keras-onnx icon indicating copy to clipboard operation
keras-onnx copied to clipboard

IndexError: list index out of range

Open alexiycv opened this issue 5 years ago • 0 comments

lipreadnet_small.zip

SharedScreenshot IndexError Traceback (most recent call last) in ----> 1 onnx_model = keras2onnx.convert_keras(model, model.name)

D:\anaconda3\lib\site-packages\keras2onnx\main.py in convert_keras(model, name, doc_string, target_opset, initial_types, channel_first_inputs, debug_mode, custom_op_conversions) 100 parse_graph_modeless(topology, tf_graph, target_opset, input_names, output_names, output_dict) 101 else: --> 102 parse_graph(topology, tf_graph, target_opset, output_names, output_dict) 103 topology.compile() 104

D:\anaconda3\lib\site-packages\keras2onnx\parser.py in parse_graph(topo, graph, target_opset, output_names, keras_node_dict) 908 return _parse_graph_core_v2( 909 graph, keras_node_dict, topo, top_level, output_names --> 910 ) if is_tf2 and is_tf_keras else _parse_graph_core( 911 graph, keras_node_dict, topo, top_level, output_names)

D:\anaconda3\lib\site-packages\keras2onnx\parser.py in _parse_graph_core_v2(graph, keras_node_dict, topology, top_scope, output_names) 782 783 if layer_info.layer and isinstance(layer_info.layer, keras.layers.TimeDistributed): --> 784 on_parsing_time_distributed_layer(graph, layer_info.nodelist, layer_info.layer, model, varset) 785 elif layer_info.layer is None or get_converter(type(layer_info.layer)) is None or
786 (isinstance(layer_info.layer, keras.layers.core.Activation) and not activation_supported):

D:\anaconda3\lib\site-packages\keras2onnx\parser.py in _on_parsing_time_distributed_layer(graph, node_list, layer, model, varset, prefix) 134 135 if isinstance(layer.layer, keras.Model): --> 136 kenode = extract_inbound_nodes(layer.layer)[0] 137 intop = varset.declare_local_operator(TYPES.Identity) 138 intop.add_input(i0_reshape)

IndexError: list index out of range

my model structure: model my model file is attached below

alexiycv avatar Nov 27 '20 07:11 alexiycv