TDA4VE-Q1: [TIDL] Onnx model with gather layer has wrong result

Hi Henry;

Thanks for the question. 

Could you provide the screenshot (or the log file) of your compiling and inferencing (run model).

Regards

Wen Li 

Hi, 

   OK. 

   In addition, I show my options and the intermediate files during the compilation process.
   onnx_cpu_output.bin is the result of cpuExecution
   tidl_output.bin is the result of tidlExecution

{
'tidl_tools_path': '/home/root/tidl_tools',
'artifacts_folder': '/home/root/output/',
'platform': 'J7',
'version': '8.6',
'tensor_bits': 8,
'debug_level': 0,
'max_num_subgraphs': 16,
'deny_list': '',
'deny_list:layer_type': '',
'deny_list:layer_name': '',
'model_type': '',
'accuracy_level': 1,
'advanced_options:calibration_frames': 1,
'advanced_options:calibration_iterations': 10,
'advanced_options:output_feature_16bit_names_list': '',
'advanced_options:params_16bit_names_list': '',
'advanced_options:mixed_precision_factor': -1,
'advanced_options:': 0,
'advanced_options:high_resolution_optimization': 0,
'advanced_options:pre_batchnorm_fold': 1,
'ti_internal_nc_flag': 1601,
'advanced_options:activation_clipping': 1,
'advanced_options:weight_clipping': 1,
'advanced_options:bias_calibration': 1,
'advanced_options:add_data_convert_ops': 3,
'advanced_options:channel_wise_quantization': 0,
'advanced_options:prequantized_model': 0,
'advanced_options:inference_mode': 0,
'advanced_options:num_cores': 1
}

log_and_file.zip

Thanks

Henry