how to add multi-inputs? or how to test the benchmark on multi inputs? about mace-models

[lee-bin]

Does your model have multiple input nodes? Or you want to validate the model with different input data? You can paste the yaml file if it helps to clarify.

[peiwenYe]

Thank you for your reply very much. Here is my yaml, I just want to test multi input with the same input format. If I only put one input like "validation_inputs_data: - /home/ypw/speech/kaldi-onnx-master/mace_input.npy", It works.
library_name: aishell2
target_abis: [aarch64-linux-gnu]
model_graph_format: file
model_data_format: file
models:
aishell2:
platform: onnx
model_file_path: /home/ypw/speech/kaldi-onnx-master/text.onnx
model_sha256_checksum: 8d60e44c128523a82c82f9d592912fa9fe8923ad81443b66c9932f92d0d1dc3f
subgraphs:

• input_tensors:
• input
  input_shapes:
• 1,52,40
  output_tensors:
  #- output.affine
  #- output-xent.log-softmax
• output.linear
  output_shapes:
• 1,7,4408
  backend: kaldi
  input_data_formats: NONE
  output_data_formats: NONE
  validation_inputs_data:
• [/home/ypw/speech/kaldi-onnx-master/mace_input.npy,/home/ypw/speech/kaldi-onnx-master/test_input.npy]
  validation_outputs_data:
• [/home/ypw/speech/kaldi-onnx-master/mace_output.npy,/home/ypw/speech/kaldi-onnx-master/test_output.npy]
  runtime: cpu
  limit_opencl_kernel_time: 0
  nnlib_graph_mode: 0
  obfuscate: 0

[lee-bin]

MACE only supports validation for one group of inputs (for models with multiple input nodes), you can't specify multiple groups of input data. You can validate it multiple times with different groups of input data or write your own codes to do it.

[peiwenYe]

I got it. Thank you very much!