Convolution model architecture

General Discussion
, , ,
#1

s this code a good equivalent of this image? I know that I have to tokenize the input on character level before that, I mean just the network architecture
Link to picture: https://i.stack.imgur.com/D3K8o.png

    model = tf.keras.Sequential([
        Embedding(input_dim=params["input_dim"], output_dim=16),
        Conv1D(filters=128, kernel_size=2, padding="valid", input_shape=(params["input_dim"], 16)),
        Conv1D(filters=192, kernel_size=3, padding="valid"),
        Conv1D(filters=256, kernel_size=4, padding="valid"),
        Conv1D(filters=512, kernel_size=5, padding="valid"),
        BatchNormalization(),
        Conv1D(filters=128, kernel_size=2, padding="valid"),
        Conv1D(filters=128, kernel_size=3, padding="valid"),
        Conv1D(filters=256, kernel_size=4, padding="valid"),
        Conv1D(filters=512, kernel_size=5, padding="valid"),
        GlobalAvgPool1D(),
        BatchNormalization(),
        BatchNormalization(),
        BatchNormalization(),
        Dense(params["output_dim"])
    ])
#3

Hi @Patryk_Bartkowiak you can use tf.keras.utils.plot_model which will convert a Keras model to dot format and save to a file and you can verify with the image you provided. For reference please refer this link