Configuring Dense Layers for Time Series Prediction

Hi Team,

I am trying to train a Time Series Model in KNIME AP, with the aim of deploying it to KNIME HUB in the near future.

This is what the flow looks like:

This is what my data looks like at the output table marked with red circle:

Before using KNIME I had prepared the model in my local laptop Pycharm and decided that I would like to try this dense keras model in KNIME AP first as a test:

Pic 1 shows my current workflow and the location of the Keras Dense nodes in yellow, I have ordered them as per what I learned from the community.

I need help in the following:
1. How to configure the keras input layer:
Current configuration:

2. How to configure the rest of the layers as per the requirement. Below is the last layer's configuration:

This is error I am getting currently:
ARN Keras Network Learner 3:37 Selected target columns provide more elements (73) than neurons available (1) for network target ‘dense_3_0:0’. Try removing some columns from the selection.

3. An answer to this question : Once I develop the model using the this technique I will be using the trained model it in a different workflow (local/server) for inference (let's call this inference workflow). Will the inference workflow need Python/conda to work. The training required me to install python and some dependencies and I am pretty sure my org will not have those in the HUB. So please let me know if we need to have conda for inference.

For reference this is how the learner has been configured

Somehow figured out why the error in the main thread was coming and fixed it but now i am getting this
line 232, in hash
‘Instead, use tensor.ref() as the key.’ % self)
TypeError: Tensors are unhashable. (KerasTensor(type_spec=TensorSpec(shape=(32, 72), dtype=tf.float32, name=‘input_1’), name=‘input_1’, description=“created by layer ‘input_1’”))Instead, use tensor.ref() as the key.

Please note the 72 columns is 9 columns being lagged upto 8 times. hence 9 X 8