Pytorch torch.nn.functional.avg_pool stride defalut None cause error in _infer_type

masahi · March 4, 2020, 4:57pm

Thanks, optional arguments should be handeled something like this

apache/incubator-tvm/blob/5a0f39b5481a30a2eec49e27cbc17a722bd6ee6a/python/tvm/relay/frontend/pytorch.py#L493-L504


if inputs[1]:
    axis = _infer_shape(inputs[1])
else:
    axis = None
if len(inputs) > 2 and inputs[2]:
    keepdims = int(inputs[2])
else:
    keepdims = False
if len(inputs) > 3 and inputs[3]:
    exclude = int(inputs[3])
else:
    exclude = False

Can you send a PR to fix this?