TVM returns cryptic error message when the input tensor type is incorrect

[jxx123]

Consider an input tensor like this
x = np.array(img).reshape(1, -1, 300, 300).astype(np.float32)

When the type casting is missing, e.g. x = np.array(img).reshape(1, -1, 300, 300), tvm returns interesting error msg like this (could be potential patch for tvm):

Traceback (most recent call last):

  File "src/onnx_to_relay.py", line 42, in <module>
    module.set_input(input_blob.name, tvm.nd.array(x))

  File "/home/jinyu/tvm/python/tvm/contrib/graph_runtime.py", line 158, in set_input
    self._get_input(key).copyfrom(value)

  File "/home/jinyu/tvm/python/tvm/runtime/ndarray.py", line 121, in copyfrom
    source_array.copyto(self)

  File "/home/jinyu/tvm/python/tvm/runtime/ndarray.py", line 186, in copyto
    return self._copyto(target)

  File "/home/jinyu/tvm/python/tvm/_ffi/_ctypes/ndarray.py", line 88, in _copyto
    check_call(_LIB.TVMArrayCopyFromTo(self.handle, target_nd.handle, None))

  File "/home/jinyu/tvm/python/tvm/_ffi/base.py", line 329, in check_call
    raise get_last_ffi_error()

tvm._ffi.base.TVMError: Traceback (most recent call last):
  [bt] (1) /home/jinyu/tvm/build/libtvm.so(TVMArrayCopyFromTo+0xa) [0x7f35ef51d90a]
  [bt] (0) /home/jinyu/tvm/build/libtvm.so(tvm::runtime::NDArray::CopyFromTo(DLTensor const*, DLTensor*, void*)+0x7c0) [0x7f35ef51cf9
0]
  File "../src/runtime/ndarray.cc", line 222
TVMArrayCopyFromTo: Check failed: from_size == to_size (270000 vs. 1080000) : The size must exactly match