In order for us to best help you, could you provide a full standalone example of code that gives you this error? Your code above has a number of issues that prevent it from running (missing imports, indentation problems). I’m hesitant to debug the code unless I can reproduce your error msg without having to modify the code.
Looking at your code and the error msg, my guess is that you have not aligned the batch_size correctly with the size of data you are feeding in.
For example, you create your features like _x1 = tf.Variable(tf.convert_to_tensor(feat_df[0], dtype=tf.float32))
so these variables will have the same size as the number of datapoints in the training data (150 I believe). However, you later set a batch_size of 6. So when you feed in your datapoints later on to the engine, it expects either scalar values, or vectors with six dimensions, not vectors with 150 dimensions.
Something like **{"x1": _x1[:6], "x2": _x2[:6],"x3": _x3[:6], "x4": _x4[:6]}
should work. If you want to iterate through the dataset, you can change [:6] to dynamically iterate through the training data at each step of training.
As a tip for debugging these kinds of things yourself, you can always print the output of, e.g., _x1 in the notebook and see if it’s shape/values match what you expect.