Error when quantizing model

[wangjia184]

I am encountering the following error when quantizing a model converted from tensorflow

File "/opt/homebrew/lib/python3.10/site-packages/onnxruntime/quantization/quantize.py", line 188, in quantize_static
    quantizer.quantize_model()
  File "/opt/homebrew/lib/python3.10/site-packages/onnxruntime/quantization/qdq_quantizer.py", line 157, in quantize_model
    op_quantizer.quantize()
  File "/opt/homebrew/lib/python3.10/site-packages/onnxruntime/quantization/operators/activation.py", line 106, in quantize
    if not self.quantizer.is_activation_symmetric and self.quantizer.try_replacing_upstream_output(
  File "/opt/homebrew/lib/python3.10/site-packages/onnxruntime/quantization/qdq_quantizer.py", line 184, in try_replacing_upstream_output
    self.tensors_to_quantize.remove(upstream_output_name)
ValueError: list.remove(x): x not in list

The model is firstly converted from TF2.0 Keras model into ONNX model successfully. But then quantize_static failed

def convert_to_onnx_model(file_name):
    spec = (tf.TensorSpec((1, 4, 15, 15), tf.float32, name="input"),)
    tf2onnx.convert.from_keras( mymodel, input_signature=spec, opset=17, output_path=file_name+".2")
    dr = SampleInputDataReader()
    quantize_static(
        file_name+".2",
        file_name,
        dr,
        quant_format=QuantFormat.QDQ,
        per_channel=False,
        activation_type=QuantType.QInt8,
        weight_type=QuantType.QInt8,
        optimize_model=False,
    )

It failed at this line

Further information

• Is this issue related to a specific model?
  Model name: my custom model

Here is the ONNX model converted from tensorflow model, you can check it in tools like Netron

Model opset: 17

Does anyone have an idea why it failed?

[yufenglee]

It should be same issue as #12556. Could you please upgrade to 1.13.1 and have a try?