Cannot Create N-Beat Model

[jimmylai-hk]

Based on the following code in colab, (in fact I just copy and paste the codes), I can run up to step 6.

%%time

tf.random.set_seed(42)

1. Setup N-BEATS Block layer

nbeats_block_layer = NBeatsBlock(input_size=INPUT_SIZE,
theta_size=THETA_SIZE,
horizon=HORIZON,
n_neurons=N_NEURONS,
n_layers=N_LAYERS,
name="InitialBlock")

2. Create input to stacks

stack_input = layers.Input(shape=(INPUT_SIZE), name="stack_input")

3. Create initial backcast and forecast input (backwards predictions are referred to as residuals in the paper)

residuals, forecast = nbeats_block_layer(stack_input)

Add in subtraction residual link, thank you to: #174

residuals = layers.subtract([stack_input, backcast], name=f"subtract_00")

4. Create stacks of blocks

for i, _ in enumerate(range(N_STACKS-1)): # first stack is already creted in (3)

5. Use the NBeatsBlock to calculate the backcast as well as block forecast

backcast, block_forecast = NBeatsBlock(
input_size=INPUT_SIZE,
theta_size=THETA_SIZE,
horizon=HORIZON,
n_neurons=N_NEURONS,
n_layers=N_LAYERS,
name=f"NBeatsBlock_{i}"
)(residuals) # pass it in residuals (the backcast)

6. Create the double residual stacking

residuals = layers.subtract([residuals, backcast], name=f"subtract_{i}")
forecast = layers.add([forecast, block_forecast], name=f"add_{i}")

7. Put the stack model together

model_7 = tf.keras.Model(inputs=stack_input,
outputs=forecast,
name="model_7_N-BEATS")

If I want to build the model, (step 7), I have the following errors.

alueError Traceback (most recent call last)
in ()
----> 1 get_ipython().run_cell_magic('time', '', '\ntf.random.set_seed(42)\n\n# 1. Setup N-BEATS Block layer\nnbeats_block_layer = NBeatsBlock(input_size=INPUT_SIZE,\n theta_size=THETA_SIZE,\n horizon=HORIZON,\n n_neurons=N_NEURONS,\n n_layers=N_LAYERS,\n name="InitialBlock")\n\n# 2. Create input to stacks\nstack_input = layers.Input(shape=(INPUT_SIZE), name="stack_input")\n\n# 3. Create initial backcast and forecast input (backwards predictions are referred to as residuals in the paper)\nresiduals, forecast = nbeats_block_layer(stack_input)\n# Add in subtraction residual link, thank you to: #174 \nresiduals = layers.subtract([stack_input, backcast], name=f"subtract_00") \n\n# 4. Create stacks of blocks\nfor i, _ in enumerate(range(N_STACKS-1)): # first stack is already creted in (3)\n\n # 5. Use the NBeatsBlock to calculate the backcast as well as block forecast\n backcast, block_forecast = NBeatsBlock(\n input_size=INPUT_SIZE,\n theta_size=THETA_SIZE,\n horizon=HORIZON,\n n_neurons=N_NEURONS,\n n_layers=N_LAYERS,\n name=f"NBeatsBlock_{i}"\n )(residuals) # pass it in residuals (the backcast)\n\n ...

7 frames
in time(self, line, cell, local_ns)

in ()

/usr/local/lib/python3.7/dist-packages/keras/engine/functional.py in _map_graph_network(inputs, outputs)
982 'The following previous layers '
983 'were accessed without issue: ' +
--> 984 str(layers_with_complete_input))
985 for x in tf.nest.flatten(node.outputs):
986 computable_tensors.add(id(x))

ValueError: Graph disconnected: cannot obtain value for tensor KerasTensor(type_spec=TensorSpec(shape=(None, 7), dtype=tf.float32, name='stack_input'), name='stack_input', description="created by layer 'stack_input'") at layer "subtract_00". The following previous layers were accessed without issue: []

[mrdbourke]

Hey @jimmylai-hk,

What have you tried to fix it so far? (troubleshooting is a good way to learn how to fix things)

I just pushed an update which may have been causing your error, I can confirm the code runs on my end in this notebook: https://github.com/mrdbourke/tensorflow-deep-learning/blob/main/10_time_series_forecasting_in_tensorflow.ipynb

[jimmylai-hk]

I didnt modify anything for the code. I just reopen today and found it works. I dont know why. Anyway, thx for your reply.