Diff Coverage

Diff: origin/master...HEAD, staged and unstaged changes

Total: 48 lines
Missing: 4 lines
Coverage: 91%

Source File	Diff Coverage (%)	Missing Lines
hyper_parallel/core/expert_parallel/expert_parallel.py	90.0%	811,815-816
hyper_parallel/platform/__init__.py	100%
hyper_parallel/platform/platform.py	91.7%	860
hyper_parallel/platform/torch/platform.py	100%

hyper_parallel/core/expert_parallel/expert_parallel.py


        ep_mesh = device_mesh["ep"]

        if self.async_combine:
            handle = AllToAllTokenDispatcher.combine_start(
                routed_output, ep_mesh, ctx
            )
            # pylint: disable=W0212
            module._ep_combine_handle = handle
            return handle.wait()

        return AllToAllTokenDispatcher.combine(
            module, routed_output, ep_mesh, ctx,
        )

hyper_parallel/platform/platform.py


        Raises:
            NotImplementedError: Must be implemented by platform subclasses.
        """
        raise NotImplementedError(
            "Platform subclasses must implement wait_async_tensor"
        )

    @staticmethod