Fail to run person_detect.tflite via TFLite benchmark_model

[bhbruce]

Issue type

Bug

Have you reproduced the bug with TensorFlow Nightly?

Yes

Source

source

TensorFlow version

97a794b

Current behavior?

There is a TFLite model called person_detect.tflite in tflite-micro https://github.com/tensorflow/tflite-micro/blob/main/tensorflow/lite/micro/models/person_detect.tflite.
I cannot run this model via tensorflow's benchmark_model.
It looks like all depthwise_convolutions' quantization_dim meta-data are wrong.

Command to reproduce the issue

`tools/benchmark/benchmark_model  --num_runs=3 --enable_op_profiling=true --num_threads=2 --graph=person_detect.tflite`

Relevant log output

### Error Log 

INFO: STARTING!
INFO: Log parameter values verbosely: [0]
INFO: Min num runs: [3]
INFO: Num threads: [2]
INFO: Graph: [person_detect.tflite]
INFO: Signature to run: []
INFO: Enable op profiling: [1]
INFO: #threads used for CPU inference: [2]
INFO: Loaded model person_detect.tflite
ERROR: quantized_dimension must be in range [0, 1). Was 3.
ERROR: Tensor 33 has invalid quantization parameters.
ERROR: quantized_dimension must be in range [0, 1). Was 3.
ERROR: Tensor 36 has invalid quantization parameters.
ERROR: quantized_dimension must be in range [0, 1). Was 3.
ERROR: Tensor 40 has invalid quantization parameters.
ERROR: quantized_dimension must be in range [0, 1). Was 3.
ERROR: Tensor 44 has invalid quantization parameters.
ERROR: quantized_dimension must be in range [0, 1). Was 3.
ERROR: Tensor 48 has invalid quantization parameters.
ERROR: quantized_dimension must be in range [0, 1). Was 3.
ERROR: Tensor 52 has invalid quantization parameters.
ERROR: quantized_dimension must be in range [0, 1). Was 3.
ERROR: Tensor 56 has invalid quantization parameters.
ERROR: quantized_dimension must be in range [0, 1). Was 3.
ERROR: Tensor 60 has invalid quantization parameters.
ERROR: quantized_dimension must be in range [0, 1). Was 3.
ERROR: Tensor 64 has invalid quantization parameters.
ERROR: quantized_dimension must be in range [0, 1). Was 3.
ERROR: Tensor 68 has invalid quantization parameters.
ERROR: quantized_dimension must be in range [0, 1). Was 3.
ERROR: Tensor 72 has invalid quantization parameters.
ERROR: quantized_dimension must be in range [0, 1). Was 3.
ERROR: Tensor 76 has invalid quantization parameters.
ERROR: quantized_dimension must be in range [0, 1). Was 3.
ERROR: Tensor 80 has invalid quantization parameters.
ERROR: quantized_dimension must be in range [0, 1). Was 3.
ERROR: Tensor 84 has invalid quantization parameters.
ERROR: Failed to initialize the interpreter
ERROR: Benchmarking failed.

[sawantkumar]

hi @bhbruce ,

I will benchmark the model and will get back to you.

[sawantkumar]

hi @bhbruce ,

I replicated the issue and i got the same error. So its probably a bug in the model's metadata . I will raise a PR to get it fixed. Thank you

[bhbruce]

Hi @sawantkumar
Appreciate for confirming it. If you have a PR to fix it, please share it with me.
Thanks.