Have a question? Connect with the community at the TensorFlow Forum Visit Forum

tflite_model_maker.recommendation.CnnSpec

Creates Recommendation CNN model spec. See also: tflite_model_maker.recommendation.ModelSpec.

encoder_type str, encoder type. One of ('bow', 'cnn', 'rnn').
context_embedding_dim int, dimension of context embedding layer.
label_embedding_dim int, dimension of label embedding layer.
item_vocab_size int, the size of items to be predict.
num_predictions int, the number of top-K predictions in the output.
hidden_layer_dim_ratios list of float, number of units in hidden layers specified by ratios. default: [1.0, 0.5, 0.25].
conv_num_filter_ratios list of int, for 'cnn', Conv1D layers' filter ratios based on context_embedding_dim.
conv_kernel_size int, for 'rnn', Conv1D layers' kernel size.
lstm_num_units int, for 'rnn', LSTM layer's unit number.
eval_top_k list of int, evaluation metrics for a list of top k.
batch_size int, default batch size.