Welcome to our Support Center

Troubleshooting

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

CodeVIInformation
5001check_bidirectional_outputThe 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).
5002check_bidirectional_return_sequencesThe layers sent in the “layer” and “backward_layer” parameters of the Bidirectional layer must have the same value for the return_sequences parameter.
5003check_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 – 5027check_attentionThe entries in the Attention layer do not have the correct input shape.
5004 – 5005 – 5026check_additive_attentionThe entries in theΒ AdditiveAttention layer do not have the correct input shape.
5004 – 5026check_multiheadattentionThe entries in the MultiHeadAttention layer do not have the correct input shape.
5006check_input_same_indexThe index selected when formatting the data at the input of the forward is already defined.
5007check_input_shapeThe 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.
5008check_nb_inputThe number of inputs sent to the forward does not match the number of inputs in the model.
5009check_nb_y_trueThe 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.
5010warning_input_shapeAn input already exists, so the input shape you defined in the “in/out param” parameter is not taken into account.
5011check_output_same_indexThe index selected when formatting the data sent to the loss (y_true) is already defined.
5012check_shapeThe inputs of the operand (Add or Average or Multiply or Substract) do not have the same shape.
5013check_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).

5014check_shape_set_inputThe dimensions of the input array provided does not correspond to those expected.
5015check_shape_set_weightThe dimensions of the weight array provided does not correspond to those expected.
5016check_shape_y_trueThe dimensions of the y_true array provided does not correspond to those expected.
5018warning_output_indexThe output order specified is out of bound (x < 0 or x > nb_output).
5019warning_input_nameThe layer name specified isn’t an input of model.
5020check_output_same_nameThe name specified when formatting the data sent to the loss (y_true) is already defined.
5021warning_input_indexThe input order specified is out of bound (x < 0 or x > nb_input).
5022display_name_errorThe name specified doesn’t exist in the graph.
5023 – 5039check_gpu_layerOne or more layers are not available in CUDA version.
5024display_dim_errorThe layer does not support the given dimension.
5028check_dimThe input is incompatible with the layer.
5029check_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).

5030check_graphs_shapeThe output shape of the first graph is incompatible with the input shape of the second.
5031warning_y_true_nameThe name shown is not a driveable output of the model.
5032 – 5033check_pool_shapeThe kernel size or the stride does not have the right shape or right values.
5034check_filter_unitsThe value assigned to the “n_filters” parameter must be strictly positive.
5035check_embedding_input_output_dimThe Embedding layer must have strictly positive the “input_dim” and “output_dim” parameters value.
5036check_rnn_gpu_integrationThe “activation” parameter(s) of the cells sent to the RNN layer is not compatible with the CuDNN version.
5037check_fit_metric_bestThe metric selected in the “Comparison Metric” parameter of the “Fit” function is not in the ‘metrics_array’.
5038check_multi_input_batchThe batch size is not the same for every inputs of operand layer (Add or Average or Multiply or Substract).
5040 – 5042check_gpu_readyYour gpu is not ready. You must have a CUDA compatible graphics card or install CUDA. You can use our installer to start the installation.
5043warning_attention_scaleThe AdditiveAttention layer has no weight because “use_scale” == False.
5044check_indexThe index specified doesn’t exist in the graph.
5050Licence not installedThe licence is not installed so toolkit could not run.
5051Licence expiredLicence expired so toolkit could not run.

 

Table of Contents

Table of Contents

Index