HinSAGE gives list index out of range as soon as I make the layer size a 2D array

[akshatmalik]

Describe the bug

As soon as I make the layer sizes and num_samples a 2D array, the HinSage functions gives the following error

To Reproduce

Steps to reproduce the behavior:

1. Go to '...'
2. Execute '...'
3. See error

Observed behavior

Traceback (most recent call last):
  File "/Users/maalik/PycharmProjects/GraphDefectPredictionML/main.py", line 171, in <module>
    activations=["relu", "relu"],
  File "/Users/maalik/opt/anaconda3/envs/GraphDefectPredictionML/lib/python3.6/site-packages/stellargraph/layer/hinsage.py", line 405, in __init__
    for layer, dim in enumerate([self.input_dims] + layer_sizes)
  File "/Users/maalik/opt/anaconda3/envs/GraphDefectPredictionML/lib/python3.6/site-packages/stellargraph/layer/hinsage.py", line 405, in <listcomp>
    for layer, dim in enumerate([self.input_dims] + layer_sizes)
IndexError: list index out of range

Expected behavior

Should not give an error, should make the model.

Environment

Operating system: for example: macOS, Ubuntu

Python version: for example: 3.7.2

Package versions: for example: stellargraph==0.8.3, tensorflow==2.0.0

The following Python code can automatically compute the data for this section (for example: run the code in a Jupyter notebook cell, or, on macOS, copy the code and run pbpaste | python - in a terminal):

    sgi = StellarGraph.from_networkx(nx_g, node_features="features", edge_weight_attr="features")

    print(sgi.info())

    train_nodes, test_nodes = model_selection.train_test_split(df, train_size=0.8, test_size=None, stratify=df["buggy"])
    all_labels = df["buggy"]
    train_labels = all_labels.loc[train_nodes.index]
    test_labels = all_labels.loc[test_nodes.index]

    generator = HinSAGENodeGenerator(sgi, batch_size=200, num_samples=[3, 4], head_node_type ="commit")
    train_gen = generator.flow(train_nodes.index, train_labels, shuffle=True)
    test_gen = generator.flow(test_nodes.index, test_labels)

    model = HinSAGE(
        layer_sizes=[10, 20],
        generator=generator,
        dropout=0.5,
        bias=True,
        normalize="l2",
        activations=["relu", "relu"],
    )

Additional context

Add any other context about the problem here. For example:

Data used:
StellarDiGraph: Directed multigraph
Nodes: 17146, Edges: 59737

Node types:
commit: [12450]
Features: float32 vector, length 4
Edge types: commit-default->person, commit-modified->file, commit-parent->commit
file: [4418]
Features: none
Edge types: none
person: [278]
Features: none
Edge types: none

Edge types:
commit-modified->file: [43419]
Weights: range=[0, 9070], mean=34.0921, std=139.13
Features: none
commit-parent->commit: [8159]
Weights: all 1 (default)
Features: none
commit-default->person: [8159]
Weights: all 1 (default)
Features: none