Error of input shape {0,32,1,1} when infer with generated onnx model

[DancingJane]

I have already converted tensorflow model in ckpt into onnx, there is no error.
However, this error occurs when inferring with the generated onnx file

Traceback (most recent call last):
File "tf2onnx.py", line 438, in
onnx_infer(outfile)
File "tf2onnx.py", line 206, in onnx_infer
outputs = sess.run([], {sess.get_inputs()[0].name:input_img, sess.get_inputs()[1].name:input_fp, sess.get_inputs()[2].name:input_agl, sess.get_inputs()[3].name:phase_train})
File "/Users/jane/opt/anaconda3/envs/python37/lib/python3.7/site-packages/onnxruntime/capi/onnxruntime_inference_collection.py", line 200, in run
return self._sess.run(output_names, input_feed, run_options)
onnxruntime.capi.onnxruntime_pybind11_state.RuntimeException: [ONNXRuntimeError] : 6 : RUNTIME_EXCEPTION : Non-zero status code returned while running Reshape node. Name:'Transpose__413' Status Message: /private/var/folders/sy/f16zz6x50xz3113nwtb9bvq00000gp/T/abs_f6klxfmwxo/croots/recipe/onnxruntime_1664948961123/work/onnxruntime/core/providers/cpu/tensor/reshape_helper.h:41 onnxruntime::ReshapeHelper::ReshapeHelper(const onnxruntime::TensorShape &, onnxruntime::TensorShapeVector &, bool) gsl::narrow_cast<int64_t>(input_shape.Size()) == size was false. The input tensor cannot be reshaped to the requested shape. Input shape:{0,32,1,1}, requested shape:{1,32,1,1}

the tf2onnx command is as follows:
python -m tf2onnx.convert
--checkpoint "./L.meta"
--inputs "inputs/input_img:0[1,48,64,3],inputs/input_fp:0[1,48,64,12],inputs/input_ang:0[1,2],model_control/phase_train:0"
--outputs "tower_1/warping_model/apply_lcm/Add:0"
--output "./onnx/onnxModel_direct_tf15_one_output_dynamic_4.onnx"
--opset 17

the inference code is as follows:
model = onnx.load(onnx_dir)
onnx.checker.check_model(model)
sess = onnxruntime.InferenceSession(onnx_dir)
input_img_shape = [1, 48, 64, 3] # Example shape, replace with the actual shape
input_img = np.random.normal(size=input_img_shape).astype(np.float32)
input_fp_shape = [1, 48, 64, 12]
input_fp = np.random.normal(size=input_fp_shape).astype(np.float32)
input_agl_shape = [1,2]
input_agl = np.random.normal(size=input_agl_shape).astype(np.float32)
phase_train = np.array(False, dtype=bool)
# Run the model (perform inference)
outputs = sess.run([sess.get_outputs()[0].name], {sess.get_inputs()[0].name:input_img, sess.get_inputs()[1].name:input_fp, sess.get_inputs()[2].name:input_agl, sess.get_inputs()[3].name:phase_train})

System information

• OS Platform and Distribution (macOS 12.1):
• TensorFlow Version: 15.0
• Python version:3.7.15
• ONNX version 1.13.0
• ONNXRuntime version :1.12.1
• tf2onnx: 1.16.1

Thank you!

[fatcat-z]

Looks like the inputs or Reshape op was not set correctly.

Is it possible for your share the model code or a code snippet which can repro this issue so I can take a debug in my local environment? Thanks.