HeteroFL baseline

[msck72]

Issue

Description

Implementation of HeteroFL baseline in flower

Related issues/PRs

Proposal

Explanation

Checklist

• Implement proposed change
• Write tests
• Update documentation
• Update changelog
• Make CI checks pass
• Ping maintainers on Slack (channel #contributions)

Any other comments?

[jafermarq]

Hi @msck72,

I have completed a first review of your baseline. There are a few code suggestions, mostly minor, but that would make your code a bit better and more similar to the other baselines. In summary my changes:

• Update pyproject.toml to the latest version of flower and python, and also install pytorch/torchvision.
• Because poetry install builds a heterofl python package, it is preferable if you execute your code like this python -m heterofl.main <options> after activating your environment via poetry shell. In order for this to work, you'll need to adjust the imports in your python files (so they follow the structure: import heterofl.<your_file> ...)

After making the changes, i was able to run your code with python -m heterofl.main but then I encountered a bug when the strategy evaluates the initial parameters.

[2023-10-02 15:40:25,663][flwr][INFO] - Evaluating initial parameters
start of testing
end of stat, time taken = 91.22248697280884
ERROR flwr 2023-10-02 15:41:57,147 | app.py:295 | tensors used as indices must be long, byte or bool tensors
[2023-10-02 15:41:57,147][flwr][ERROR] - tensors used as indices must be long, byte or bool tensors
ERROR flwr 2023-10-02 15:41:57,147 | app.py:296 | Traceback (most recent call last):
  File "/home/javier/.cache/pypoetry/virtualenvs/heterofl-n4YayD5Q-py3.10/lib/python3.10/site-packages/flwr/simulation/app.py", line 290, in start_simulation
    hist = run_fl(
  File "/home/javier/.cache/pypoetry/virtualenvs/heterofl-n4YayD5Q-py3.10/lib/python3.10/site-packages/flwr/server/app.py", line 224, in run_fl
    hist = server.fit(num_rounds=config.num_rounds, timeout=config.round_timeout)
  File "/home/javier/.cache/pypoetry/virtualenvs/heterofl-n4YayD5Q-py3.10/lib/python3.10/site-packages/flwr/server/server.py", line 92, in fit
    res = self.strategy.evaluate(0, parameters=self.parameters)
  File "/home/javier/projects/sor/heterofl/baselines/heterofl/heterofl/strategy.py", line 369, in evaluate
    eval_res = self.evaluate_fn(server_round, parameters_ndarrays, {})
  File "/home/javier/projects/sor/heterofl/baselines/heterofl/heterofl/server.py", line 78, in evaluate
    client_test_res = test(
  File "/home/javier/projects/sor/heterofl/baselines/heterofl/heterofl/models.py", line 501, in test
    output = model(input_dict)
  File "/home/javier/.cache/pypoetry/virtualenvs/heterofl-n4YayD5Q-py3.10/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1194, in _call_impl
    return forward_call(*input, **kwargs)
  File "/home/javier/projects/sor/heterofl/baselines/heterofl/heterofl/models.py", line 337, in forward
    label_mask[input_dict["label_split"]] = 1
IndexError: tensors used as indices must be long, byte or bool tensors

Please ping me when you have a chance to make the suggested changes (and mark them as resolved when you do). If something is not clear, happy to discuss further -- or ping me on Slack.

[jafermarq]

Hi @msck72, Great job implementing HeteroFL!