optima2005 commented on issue #4418: [RUNTIME] Add cudnn conv3d URL: https://github.com/apache/incubator-tvm/pull/4418#issuecomment-559451369 _get_np_int32_array_handle works for conv_output_shape and conv_find_algo, but not for 'conv_forward', the problem is in the code snippet. If I changed it to ``` return _api.extern( oshape, [x, w], lambda ins, outs: _intrin.call_packed( "tvm.contrib.cudnn.conv.forward", conv_mode, tensor_format, algo, dims, _get_np_int32_array_handle(pad), _get_np_int32_array_handle(stride), _get_np_int32_array_handle(dilation), ins[0], ins[1], outs[0], conv_dtype), name="y") ``` where pad, stride, and dilation are np.array(..., dtype=np.int32), I got this error, ``` value = c_void_p(69268061568800) def convert_to_node(value): """Convert a python value to corresponding node type. ... ... ... ... if isinstance(value, string_types): return _api_internal._str(value) if isinstance(value, (list, tuple)): value = [convert_to_node(x) for x in value] return _api_internal._Array(*value) if isinstance(value, dict): vlist = [] for item in value.items(): if (not isinstance(item[0], _CLASS_NODE_BASE) and not isinstance(item[0], string_types)): raise ValueError("key of map must already been a container type") vlist.append(item[0]) vlist.append(convert_to_node(item[1])) return _api_internal._Map(*vlist) if isinstance(value, NodeGeneric): return value.asnode() if value is None: return None > raise ValueError("don't know how to convert type %s to node" % type(value)) E ValueError: don't know how to convert type <class 'ctypes.c_void_p'> to node python/tvm/_ffi/node_generic.py:95: ValueError ```
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services
