Hi,
I'm trying to compile an ONNX model using TIDL tools 08_05_00_00 (see the attached structure) and got the following error:
tidl_tools_path = /opt/tidl_tools
artifacts_folder = /data/michlit/work_data/tidl_convert/compiled_models/fvp_standalone_shared_conv_opset_11_08_05_rnd_1_iter_3
tidl_tensor_bits = 16
debug_level = 3
num_tidl_subgraphs = 16
tidl_denylist =
tidl_denylist_layer_name =
tidl_denylist_layer_type =
model_type =
tidl_calibration_accuracy_level = 7
tidl_calibration_options:num_frames_calibration = 1
tidl_calibration_options:bias_calibration_iterations = 3
mixed_precision_factor = -1.000000
model_group_id = 0
power_of_2_quantization = 2
enable_high_resolution_optimization = 0
pre_batchnorm_fold = 1
add_data_convert_ops = 0
output_feature_16bit_names_list =
m_params_16bit_names_list =
reserved_compile_constraints_flag = 1601
ti_internal_reserved_1 =
****** WARNING : Network not identified as Object Detection network : (1) Ignore if network is not Object Detection network (2) If network is Object Detection network, please specify "model_type":"OD" as part of OSRT compilation options******
Supported TIDL layer type --- Identity -- Identity_27
Supported TIDL layer type --- MaxPool -- /layers.0/layers.0.0/MaxPool
Supported TIDL layer type --- Conv -- /layers.0/layers.0.1/Conv
Supported TIDL layer type --- Relu -- /layers.0/layers.0.3/Relu
Supported TIDL layer type --- Conv -- /layers.0/layers.0.4/Conv
Supported TIDL layer type --- Relu -- /layers.0/layers.0.6/Relu
Supported TIDL layer type --- Conv -- /layers.0/layers.0.4_1/Conv
Supported TIDL layer type --- Relu -- /layers.0/layers.0.9/Relu
Unsupported (TIDL check) TIDL layer type --- MaxPool
Supported TIDL layer type --- Conv -- /layers.1/layers.1.1/Conv
Supported TIDL layer type --- Relu -- /layers.1/layers.1.3/Relu
Supported TIDL layer type --- Conv -- /layers.1/layers.1.4/Conv
Supported TIDL layer type --- Relu -- /layers.1/layers.1.6/Relu
Supported TIDL layer type --- Conv -- /layers.1/layers.1.4_1/Conv
Supported TIDL layer type --- Relu -- /layers.1/layers.1.9/Relu
Unsupported (TIDL check) TIDL layer type --- MaxPool
Supported TIDL layer type --- Conv -- /layers.2/layers.2.1/Conv
Supported TIDL layer type --- Relu -- /layers.2/layers.2.3/Relu
Supported TIDL layer type --- Conv -- /layers.2/layers.2.4/Conv
Supported TIDL layer type --- Relu -- /layers.2/layers.2.6/Relu
Supported TIDL layer type --- Conv -- /layers.2/layers.2.4_1/Conv
Supported TIDL layer type --- Relu -- /layers.2/layers.2.9/Relu
Preliminary subgraphs created = 3
Final number of subgraphs created are : 3, - Offloaded Nodes - 20, Total Nodes - 22
ALLOWLISTING : POOLING layer : Number of input dimensions must be 4 -- file info - tidl_import_common_model_check.cpp , TIDL_checkPoolingTensorProperties , 459
ALLOWLISTING : POOLING layer : Number of input dimensions must be 4 -- file info - tidl_import_common_model_check.cpp , TIDL_checkPoolingTensorProperties , 459
Running runtimes graphviz - /opt/tidl_tools/tidl_graphVisualiser_runtimes.out /data/michlit/work_data/tidl_convert/compiled_models/fvp_standalone_shared_conv_opset_11_08_05_rnd_1_iter_3/allowedNode.txt /data/michlit/work_data/tidl_convert/compiled_models/fvp_standalone_shared_conv_opset_11_08_05_rnd_1_iter_3/tempDir/graphvizInfo.txt /data/michlit/work_data/tidl_convert/compiled_models/fvp_standalone_shared_conv_opset_11_08_05_rnd_1_iter_3/tempDir/runtimes_visualization.svg
Traceback (most recent call last):
File "/data/michlit/code/Perception/autonomy/apps/lane_detection/tidl/tidl_convert.py", line 399, in <module>
main()
File "/data/michlit/code/Perception/autonomy/apps/lane_detection/tidl/tidl_convert.py", line 384, in main
input_shape=input_shape, calibration_frames=calibration_frames)
File "/data/michlit/code/Perception/autonomy/apps/lane_detection/tidl/tidl_convert.py", line 156, in compile_model
compile_options, {}], sess_options=so)
File "/usr/local/lib/python3.6/dist-packages/onnxruntime/capi/onnxruntime_inference_collection.py", line 283, in __init__
self._create_inference_session(providers, provider_options)
File "/usr/local/lib/python3.6/dist-packages/onnxruntime/capi/onnxruntime_inference_collection.py", line 315, in _create_inference_session
sess.initialize_session(providers, provider_options)
onnxruntime.capi.onnxruntime_pybind11_state.Fail: [ONNXRuntimeError] : 1 : FAIL : This is an invalid model. Error: Duplicate definition of name (feature_f1).
There is no duplicate names in the model, "feature_f1" is one of the model outputs.
The compilation options I pass:
{
'tidl_tools_path': '/opt/tidl_tools',
'artifacts_folder': <path/to/artifact/folder>,
'tensor_bits': 16,
'accuracy_level': 1,
'advanced_options:calibration_frames': 1,
'advanced_options:calibration_iterations': 3,
'debug_level': 3,
'deny_list': '',
'max_num_subgraphs': 16
}
The other question for the same model - why does the TIDL compiler issues warnings on 2 of 3 MaxPool layers - the input dimensions for all 3 MaxPool layers is 4:
ALLOWLISTING : POOLING layer : Number of input dimensions must be 4 -- file info - tidl_import_common_model_check.cpp , TIDL_checkPoolingTensorProperties , 459
Please advice.
Thanks.
