Unsupport the high dimension inputs that larger than 5.

[hailong23-jin]

It raises an assert error in the following code when our input shape is [B, C, H, W, h, w], whose dimension is larger than 5. Can I directly modify 5 to 6? Does it effect the final results?

fvcore/fvcore/nn/jit_handles.py

Lines 236 to 246 in b905b9b

	def norm_flop_jit(inputs: List[Any], outputs: List[Any]) -> Number:
	"""
	Count flops for norm layers.
	"""
	# Inputs[0] contains the shape of the input.
	input_shape = get_shape(inputs[0])
	has_affine = get_shape(inputs[affine_arg_index]) is not None
	assert 2 <= len(input_shape) <= 5, input_shape
	# 5 is just a rough estimate
	flop = prod(input_shape) * (5 if has_affine else 4)
	return flop