Incompatible attr in node elemwise_add0 at 1-th input: expected [1,180], got [180]

[mnboos]

See: https://discuss.tvm.ai/t/nnvm-base-nnvmerror-cannot-find-argument-auto-pad/1644/3

This error occurs when I try to compile an ONNX model. I'm currently clarifying if I'm allowed to share the model or if it's confidential.

Traceback (most recent call last):
  File "/home/martin/Dev/s-tvm/compile_model.py", line 189, in <module>
    compile_model(model_name=model_name, model_path=model_path, output_dir=output_dir)
  File "/home/martin/Dev/s-tvm/compile_model.py", line 93, in compile_model
    dtype=input_dtype_dict)
  File "/home/martin/.local/lib/python3.6/site-packages/nnvm-0.8.0-py3.6.egg/nnvm/compiler/build_module.py", line 270, in build
    ishape, _ = graph_util.infer_shape(graph, **shape)
  File "/home/martin/.local/lib/python3.6/site-packages/nnvm-0.8.0-py3.6.egg/nnvm/compiler/graph_util.py", line 31, in infer_shape
    graph = graph.apply("InferShape")
  File "/home/martin/.local/lib/python3.6/site-packages/nnvm-0.8.0-py3.6.egg/nnvm/graph.py", line 234, in apply
    check_call(_LIB.NNGraphApplyPasses(self.handle, npass, cpass, ctypes.byref(ghandle)))
  File "/home/martin/.local/lib/python3.6/site-packages/nnvm-0.8.0-py3.6.egg/nnvm/_base.py", line 75, in check_call
    raise NNVMError(py_str(_LIB.NNGetLastError()))
nnvm._base.NNVMError: Error in operator elemwise_add0: [07:37:27] /home/martin/Dev/tvm/nnvm/src/top/nn/../elemwise_op_common.h:38: Check failed: assign(&dattr, (*vec)[i]) Incompatible attr in node elemwise_add0 at 1-th input: expected [1,180], got [180]

cc @srkreddy1238

[mnboos]

See the attached model for reproducing the errors:

model.zip

[tqchen]

@srkreddy1238 can you followup on this?

[mnboos]

Any update on this?

[mnboos]

@tqchen I would appreciate some feedback.

[srkreddy1238]

@mnboos some how I missed it for a while. Will check it today

[srkreddy1238]

@mnboos can you tell me the graph input shape (i.e shape for conv2d_1_input_0)?

[mnboos]

According to https://lutzroeder.github.io/netron/ it's float32[1,101,161,1]

[srkreddy1238]

@mnboos
#2850 work fine for this model.
BTW, I checked with relay instead of NNVM. I will back port the same to NNVM later.

Here is the sample code

import onnx
import numpy as np
import tvm
import tvm.relay as relay

onnx_model = onnx.load('speech_rand.onnx')
target = 'llvm'

input_name = 'conv2d_1_input_0'
shape_dict = {input_name: [1, 101, 161, 1]}
dtype_dict = {input_name: 'float32'}
sym, params = relay.frontend.from_onnx(onnx_model, shape_dict, dtype=dtype_dict)

with relay.build_config(opt_level=3):
    graph, lib, params = relay.build(sym, target=target, params=params)

x = np.random.uniform(size=(1, 101, 161, 1)).astype('float32')

from tvm.contrib import graph_runtime
dtype = 'float32'
m = graph_runtime.create(graph, lib, tvm.cpu(0))
# set inputs
m.set_input(input_name, tvm.nd.array(x.astype(dtype)))
m.set_input(**params)
# execute
m.run()
# get outputs
tvm_output = m.get_output(0)

print("Out Shape:", tvm_output.shape)

[mnboos]

Thanks for the fix to both of you!