65d4d10a980fa35c383e2a0c583224d5d2ee13b6,python/src/nnabla/core/variable_batch_size.py,,variable_batch_size,#Any#,18

Before Change

        b = network.variables[pf.inputs[0]].shape[0]
        if expect_batch_size is None:
            expect_batch_size = b
        if b != expect_batch_size:
            raise ValueError("Variable "{}" has different batch size {} (expected {})".format(
                pf.inputs[0], b, expect_batch_size))

        if pf.type == "Reshape":
            arg_shape = pf.args["shape"]
            assert (arg_shape[0] == expect_batch_size)
            pf.args["shape"] = [-1] + arg_shape[1:]

After Change

        if "base_axis" in pf.args and pf.args["base_axis"] != 0:
            base_axis = pf.args["base_axis"]
            for pv_name in pf.inputs:
                if pv_name in network.variables:
                    shape = network.variables[pv_name].shape
                    expect_batch_size = np.prod(shape[:base_axis])
                    break

    if expect_batch_size is None:
        return

    // Dim 0 of shape expects to be batch size,

Instances