Function run

examples/benchmark/sparse_vs_dense_tensors.py:62–102

(sparse)

def run(sparse):
    n_classes = 10
    proportions = torch.zeros((500, n_classes), device=device)
    rates = torch.zeros((500, n_classes), device=device)
    assignments = -torch.ones(500, device=device)
    spike_record = []
    for _ in range(5):
        tmp = torch.zeros(shape, device=device)
        spike_record.append(tmp.to_sparse() if sparse else tmp)

    spike_record_idx = 0

    delta = 0
    for _ in range(10):
        start = time.perf_counter()
        label_tensor = torch.randint(0, n_classes, (n_classes,), device=device)
        spike_record_tensor = torch.cat(spike_record, dim=0)
        all_activity(
            spikes=spike_record_tensor, assignments=assignments, n_labels=n_classes
        )
        proportion_weighting(
            spikes=spike_record_tensor,
            assignments=assignments,
            proportions=proportions,
            n_labels=n_classes,
        )

        assignments, proportions, rates = assign_labels(
            spikes=spike_record_tensor,
            labels=label_tensor,
            n_labels=n_classes,
            rates=rates,
        )
        delta += time.perf_counter() - start
        spike_record[spike_record_idx] = create_spikes_tensor(
            0.03, sparse, return_memory_usage=False
        )
        spike_record_idx += 1
        if spike_record_idx == len(spike_record):
            spike_record_idx = 0
    return round(delta, 1)


def runtime_benchmark():