MNIST Tutorial hangs on mnist_clf.run on macOS

[p-i-]

In lava/tutorials/end_to_end/tutorial01_mnist_digit_classification.ipynb:

    # Run each image-inference for fixed number of steps
    mnist_clf.run(
        condition=RunSteps(num_steps=num_steps_per_image),
        run_cfg=Loihi1SimCfg(select_sub_proc_model=True,
                             select_tag='fixed_pt'))

... hangs.

#312 references exactly the same line of code, though states different symptoms. I've nevertheless implemented the suggested fix (and it doesn't make any difference).

Two students reported fail on macOS. Two on macOS using miniconda reported success. Students on Windows report success.

It seems there's a bug in the repo!

[p-i-]

I put debug-statements in lava/src/lava/magma/runtime/runtime.py:

class Runtime:
    def initialize(self, node_cfg_idx: int = 0):
        """Initializes the runtime"""
        self._build_watchdog_manager()
        print('🔸1a')
        self._build_message_infrastructure()
        print('🔸1b')
        self._build_channels()
        print('🔸1c')
        self._build_sync_channels()
        print('🔸1d')
        self._build_processes()
        print('🔸1e')
        self._build_runtime_services()
        print('🔸1f')
        self._start_ports()
        print('🔸1g')

        # Check if any exception was thrown
        for i, q in enumerate(self.exception_q):
            print(f'🟠{i}')
            e = q.get()
            if e:
                print('🟠 🟠')
                print(str(e), e.trace)
                raise(e)
        print('🟢')

Output from notebook cell:

🔸1a
🔸1b
🔸1c
🔸1d
🔸1e
objc[9426]: +[__NSCFConstantString initialize] may have been in progress in another thread when fork() was called.
objc[9426]: +[__NSCFConstantString initialize] may have been in progress in another thread when fork() was called. We cannot safely call it or ignore it in the fork() child process. Crashing instead. Set a breakpoint on objc_initializeAfterForkError to debug.
🔸1f
🔸1g
🟠0
🟠1
🟠2
🟠3
🟠4
🟠5
🟠6
🟠7