Welcome to our Support Center
Updated

Troubleshooting

This section presents the different errors that can appear when using the library.

Code VI Information
5001 check_bidirectional_output The layers sent in the “layer” and “backward_layer” parameters of the Bidirectional layer must have the same output shape, unless the value of the merge_mode parameter is “concat”. For layers to have the same output shape, the same value must be set in the “units” parameter (parameter of the layer sent to the bidirectional layer).
5002 check_bidirectional_return_sequences The layers sent in the “layer” and “backward_layer” parameters of the Bidirectional layer must have the same value for the return_sequences parameter.
5003 check_conv

The convolution is not feasible because the shape of the input is not large enough.
You have three solutions:

  • Enlarge the input size
  • Reduce the size of the kernel
  • Set the “padding” parameter of the layer to “same” (padding will be added in such a way as to convolve in all cases)
5004 – 5026 – 5027 check_attention The entries in the Attention layer do not have the correct input shape.
5004 – 5005 – 5026 check_additive_attention The entries in theΒ AdditiveAttention layer do not have the correct input shape.
5004 – 5026 check_multiheadattention The entries in the MultiHeadAttention layer do not have the correct input shape.
5006 check_input_same_index The index selected when formatting the data at the input of the forward is already defined.
5007 check_input_shape The input layer does not exist. Add the Input layer at the beginning of the model or use the “in/out param” parameter of the layer that starts your model.
5008 check_nb_input The number of inputs sent to the forward does not match the number of inputs in the model.
5009 check_nb_y_true The number of outputs (y_true) sent to the loss does not correspond to the number of trainable outputs of the model.
Caution : if your forward is in training mode it will not allow loss, if you want metric loss, use metric functionalities.
5010 warning_input_shape An input already exists, so the input shape you defined in the “in/out param” parameter is not taken into account.
5011 check_output_same_index The index selected when formatting the data sent to the loss (y_true) is already defined.
5012 check_shape The inputs of the operand (Add or Average or Multiply or Substract) do not have the same shape.
5013 check_shape_concat

The entries of the Concatenate layer do not have the same shape.

Note that the size of the axis on which we concatenate can be different.

Example: if we have an entry with shape (10, 2, 3), another with shape (10, 2, 4) and the parameter “axis” of the layer is 1, then the error occurs because when checking the shape, the layer considers that one entry has a shape of (10, 3) and the other one (10, 4). If “axis” is 2, then there is no error because during the verification of the shape, the layer considers that the entries have a shape of (10, 2).
5014 check_shape_set_input The dimensions of the input array provided does not correspond to those expected.
5015 check_shape_set_weight The dimensions of the weight array provided does not correspond to those expected.
5016 check_shape_y_true The dimensions of the y_true array provided does not correspond to those expected.
5018 warning_output_index The output order specified is out of bound (x < 0 or x > nb_output).
5019 warning_input_name The layer name specified isn’t an input of model.
5020 check_output_same_name The name specified when formatting the data sent to the loss (y_true) is already defined.
5021 warning_input_index The input order specified is out of bound (x < 0 or x > nb_input).
5022 display_name_error The name specified doesn’t exist in the graph.
5023 – 5039 check_gpu_layer One or more layers are not available in CUDA version.
5024 display_dim_error The layer does not support the given dimension.
5028 check_dim The input is incompatible with the layer.
5029 check_dim

The product of the input form is not equal to the product of the output form.

Example: for an input shape (10, 2, 3, 5) the product is equal to 300 (10 * 2 * 3 * 5) then, in output if we want 3D we will have for example a shape of (10, 3, 10).
5030 check_graphs_shape The output shape of the first graph is incompatible with the input shape of the second.
5031 warning_y_true_name The name shown is not a driveable output of the model.
5032 – 5033 check_pool_shape The kernel size or the stride does not have the right shape or right values.
5034 check_filter_units The value assigned to the “n_filters” parameter must be strictly positive.
5035 check_embedding_input_output_dim The Embedding layer must have strictly positive the “input_dim” and “output_dim” parameters value.
5036 check_rnn_gpu_integration The “activation” parameter(s) of the cells sent to the RNN layer is not compatible with the CuDNN version.
5037 check_fit_metric_best The metric selected in the “Comparison Metric” parameter of the “Fit” function is not in the ‘metrics_array’.
5038 check_multi_input_batch The batch size is not the same for every inputs of operand layer (Add or Average or Multiply or Substract).
5040 – 5042 check_gpu_ready Your gpu is not ready. You must have a CUDA compatible graphics card or install CUDA. You can use our installer to start the installation.
5043 warning_attention_scale The AdditiveAttention layer has no weight because “use_scale” == False.
5044 check_index The index specified doesn’t exist in the graph.
5050 Licence not installed The licence is not installed so toolkit could not run.
5051 Licence expired Licence expired so toolkit could not run.
5052 Runtime Error Their is a Runtime error so could not run

 

Table of Contents