When I use ONNX model to inference , such as from_onnx.py(https://docs.tvm.ai/tutorials/frontend/from_onnx.html), I find these issue :
Traceback (most recent call last):
File “from_onnx.py”, line 98, in
**tvm_output = intrp.evaluate(sym)(tvm.nd.array(x.astype(dtype)), params).asnumpy()
…
%0 = divide(%Input73, %Constant377) //
%1 = nn.conv2d(%0, %Constant321, kernel_size=[5, 5]) // an internal invariant was violdated while typechecking your program [15:47:10] /tvm/src/lang/data_layout.cc:273: Check failed: src_shape.size() == src_axis.size() (0 vs. 4) :
;
%2 = reshape(%Constant318, newshape=[8, 1, 1]) //
%3 = reshape(%2, newshape=[8]) //
%4 = expand_dims(%3, axis=1, num_newaxis=2) //
%5 = add(%1, %4) //
%6 = nn.relu(%5) //
%7 = nn.max_pool2d(%6, pool_size=[2, 2], strides=[2, 2]) // an internal invariant was violdated while typechecking your program [15:47:10] /tvm/src/relay/op/nn/pooling.cc:73: Check failed: data != nullptr:
;
%8 = nn.conv2d(%7, %Constant340, kernel_size=[5, 5]) //
%9 = reshape(%Constant346, newshape=[16, 1, 1]) //
%10 = reshape(%9, newshape=[16]) //
%11 = expand_dims(%10, axis=1, num_newaxis=2) //
%12 = add(%8, %11) //
%13 = nn.relu(%12) //
%14 = nn.max_pool2d(%13, pool_size=[3, 3], strides=[3, 3]) // an internal invariant was violdated while typechecking your program [15:47:10] /tvm/src/relay/op/nn/pooling.cc:73: Check failed: data != nullptr:
;
%15 = reshape(%14, newshape=[1, 256]) //
%16 = reshape(%Constant312, newshape=[256, 10]) //
%17 = transpose(%16, axes=[1, 0]) //
%18 = nn.dense(%15, %17, units=None) //
%19 = reshape(%Constant367, newshape=[10]) //
%20 = add(%18, %19) //
%20
I don,t know what errors happened when I use it.