/usr/bin/onnx_test_runner

/usr/bin/onnx_test_runner: invalid option -- '-'
onnx_test_runner [options...] <data_root>
Options:
	-j [models]: Specifies the number of models to run simultaneously.
	-A : Disable memory arena
	-M : Disable memory pattern
	-c [runs]: Specifies the number of Session::Run() to invoke simultaneously for each model.
	-r [repeat]: Specifies the number of times to repeat
	-m [min_tests]: Fail if fewer than this many test cases are collected (0 = no floor, the default). Guards against a silently empty/partial/truncated test tree.
	-I [inference_mode]: Use inference mode. Save the inference result and skip the output value comparison.
	-v: verbose
	-n [test_case_name]: Specifies a single test case to run.
	-e [EXECUTION_PROVIDER]: EXECUTION_PROVIDER could be 'cpu', 'cuda', 'dnnl', 'tensorrt', 'vsinpu''openvino', 'migraphx', 'acl', 'xnnpack', 'webgpu', 'nnapi', 'qnn', 'snpe' or 'coreml'. Default: 'cpu'.
	-p: Pause after launch, can attach debugger and continue
	-x: Use parallel executor, default (without -x): sequential executor.
	-d [device_id]: Specifies the device id for multi-device (e.g. GPU). The value should > 0
	-t: Specify custom relative tolerance values for output value comparison. default: 1e-5
	-a: Specify custom absolute tolerance values for output value comparison. default: 1e-5
	-C: Specify session configuration entries as key-value pairs: -C "<key1>|<value1> <key2>|<value2>"
	    Refer to onnxruntime_session_options_config_keys.h for valid keys and values.
	    [Example] -C "session.disable_cpu_ep_fallback|1 ep.context_enable|1"
	-i: Specify EP specific runtime options as key value pairs. Different runtime options available are:
	    [QNN only] [backend_type]: QNN backend type. E.g., 'cpu', 'htp'. Mutually exclusive with 'backend_path'.
	    [QNN only] [backend_path]: QNN backend path. E.g., '/folderpath/libQnnHtp.so', '/winfolderpath/QnnHtp.dll'. Mutually exclusive with 'backend_type'.
	    [QNN only] [profiling_level]: QNN profiling level, options:  'basic', 'detailed', default 'off'.
	    [QNN only] [profiling_file_path]: QNN profiling file path if ETW not enabled.
	    [QNN only] [rpc_control_latency]: QNN rpc control latency. default to 10.
	    [QNN only] [vtcm_mb]: QNN VTCM size in MB. default to 0(not set).
	    [QNN only] [htp_performance_mode]: QNN performance mode, options: 'burst', 'balanced', 'default', 'high_performance',
	    'high_power_saver', 'low_balanced', 'extreme_power_saver', 'low_power_saver', 'power_saver', 'sustained_high_performance'. Default to 'default'.
	    [QNN only] [op_packages]: QNN UDO package, allowed format:
	    op_packages|<op_type>:<op_package_path>:<interface_symbol_name>[:<target>],<op_type2>:<op_package_path2>:<interface_symbol_name2>[:<target2>].
	    [QNN only] [qnn_context_priority]: QNN context priority, options: 'low', 'normal', 'normal_high', 'high'. Default to 'normal'.
	    0 means dump the QNN context binary into separate bin file and set the path in the Onnx skeleton model.
	    [QNN only] [qnn_saver_path]: QNN Saver backend path. e.g '/folderpath/libQnnSaver.so'.
	    [QNN only] [htp_graph_finalization_optimization_mode]: QNN graph finalization optimization mode, options:
	    '0', '1', '2', '3', default is '0'.
	    [QNN only] [soc_model]: The SoC Model number. Refer to QNN SDK documentation for specific values. Defaults to '0' (unknown).
	    [QNN only] [htp_arch]: The minimum HTP architecture. The driver will use ops compatible with this architecture.
	    Options are '0', '68', '69', '73', '75', '81'. Defaults to '0' (none).
	    [QNN only] [device_id]: The ID of the device to use when setting 'htp_arch'. Defaults to '0' (for single device).
	    [QNN only] [enable_htp_fp16_precision]: Enable the HTP_FP16 precision so that the float32 model will be inferenced with fp16 precision.
	    Otherwise, it will be fp32 precision. Works for float32 model for HTP backend. Defaults to '1' (with FP16 precision.).
	    [QNN only] [offload_graph_io_quantization]: Offload graph input quantization and graph output dequantization to another EP (typically CPU EP).
	    Defaults to '0' (QNN EP handles the graph I/O quantization and dequantization).
	    [QNN only] [extended_udma]: Enable HTP extended UDMA mode for better performance on supported hardware, options:
	    '0' (disabled), '1' (enabled). Default: '0'.
	 [Usage]: -e <provider_name> -i '<key1>|<value1> <key2>|<value2>'

	 [Example] [For QNN EP] -e qnn -i "profiling_level|detailed backend_type|cpu"

	    [SNPE only] [runtime]: SNPE runtime, options: 'CPU', 'GPU', 'GPU_FLOAT16', 'DSP', 'AIP_FIXED_TF'.
	    [SNPE only] [priority]: execution priority, options: 'low', 'normal'.
	    [SNPE only] [buffer_type]: options: 'TF8', 'TF16', 'UINT8', 'FLOAT', 'ITENSOR'. default: ITENSOR'.
	    [SNPE only] [enable_init_cache]: enable SNPE init caching feature, set to 1 to enabled it. Disabled by default.
	 [Usage]: -e <provider_name> -i '<key1>|<value1> <key2>|<value2>'

	 [Example] [For SNPE EP] -e snpe -i "runtime|CPU priority|low"

	-o [optimization level]: Default is 99. Valid values are 0 (disable), 1 (basic), 2 (extended), 3 (layout), 99 (all).
		Please see onnxruntime_c_api.h (enum GraphOptimizationLevel) for the full list of all optimization levels. 	-f: Enable EP context cache generation.
	-b: Disable EP context embed mode.

	-h: help

onnxruntime version: 1.31.1