TDA4VM: segmentation fault when model import using tidl tools

Yujin Park

Prodigy 40 points

Part Number: TDA4VM

Tool/software:

Hello, I'm trying to import custom model using edgeai-tidl-tools.

My model looks llike below image.

And i got error message like this

Fullscreen

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
root@5fcd978baa80:/home/root/examples/osrt_python/tfl# python3 tflrt_delegate.py -c
Running 1 Models - ['dimension_test']
Running_Model :  dimension_test
delegate setting start
tidl_tools_path                                 = /home/root/tidl_tools 
artifacts_folder                                = ../../../custom_model/artifacts/dimension_test/ 
tidl_tensor_bits                                = 8 
debug_level                                     = 5 
num_tidl_subgraphs                              = 16 
tidl_denylist                                   = 
tidl_denylist_layer_name                        = 
tidl_denylist_layer_type                        = 
tidl_allowlist_layer_name                       = 
model_type                                      =  
tidl_calibration_accuracy_level                 = 7 
tidl_calibration_options:num_frames_calibration = 5 
tidl_calibration_options:bias_calibration_iterations = 3 
mixed_precision_factor = -1.000000 
model_group_id = 0 
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

root@5fcd978baa80:/home/root/examples/osrt_python/tfl# python3 tflrt_delegate.py -c
Running 1 Models - ['dimension_test']


Running_Model :  dimension_test
delegate setting start
tidl_tools_path                                 = /home/root/tidl_tools 
artifacts_folder                                = ../../../custom_model/artifacts/dimension_test/ 
tidl_tensor_bits                                = 8 
debug_level                                     = 5 
num_tidl_subgraphs                              = 16 
tidl_denylist                                   = 
tidl_denylist_layer_name                        = 
tidl_denylist_layer_type                        = 
tidl_allowlist_layer_name                       = 
model_type                                      =  
tidl_calibration_accuracy_level                 = 7 
tidl_calibration_options:num_frames_calibration = 5 
tidl_calibration_options:bias_calibration_iterations = 3 
mixed_precision_factor = -1.000000 
model_group_id = 0 
power_of_2_quantization                         = 3 
ONNX QDQ Enabled                                = 0 
enable_high_resolution_optimization             = 0 
pre_batchnorm_fold                              = 0 
add_data_convert_ops                            = 3 
output_feature_16bit_names_list                 =  
m_params_16bit_names_list                       =  
m_single_core_layers_names_list                 =  
Inference mode                                  = 0 
Number of cores                                 = 1 
reserved_compile_constraints_flag               = 1601 
partial_init_during_compile                     = 0 
ti_internal_reserved_1                          = 

delegate setting fin!
Model path :  ../../../custom_model/model/new_prelu_to_mul_cut5.tflite
compile mode!
========================= [Model Compilation Started] =========================

Model compilation will perform the following stages:
1. Parsing
2. Graph Optimization
3. Quantization & Calibration
4. Memory Planning

============================== [Version Summary] ==============================

-------------------------------------------------------------------------------
|          TIDL Tools Version          |              10_00_04_00             |
-------------------------------------------------------------------------------
|         C7x Firmware Version         |              10_00_02_00             |
-------------------------------------------------------------------------------

============================== [Parsing Started] ==============================

[TIDL Import] [PARSER] 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
[TIDL Import] [PARSER] SUPPORTED: Supported TIDL layer type --- 1 Tflite layer type --- 3 layer output name--- model_5/model_4/model_3/batch_normalization_132/FusedBatchNormV3;model_5/model_4/model_3/conv2d_81/BiasAdd/ReadVariableOp/resource;model_5/model_4/model_3/conv2d_81/BiasAdd;model_5/model_4/model_3/depthwise_conv2d_68/depthwise;model_5/model_4/model_3/conv2d_81/Conv2D  -- [tidl_tfLiteRtImport_core.cpp, 3090]
[TIDL Import] [PARSER] SUPPORTED: Supported TIDL layer type --- 5 Tflite layer type --- 18 layer output name--- customed_output_first  -- [tidl_tfLiteRtImport_core.cpp, 3090]
[TIDL Import] [PARSER] SUPPORTED: Supported TIDL layer type --- 5 Tflite layer type --- 18 layer output name--- customed_output_third  -- [tidl_tfLiteRtImport_core.cpp, 3090]
[TIDL Import] [PARSER] SUPPORTED: Supported TIDL layer type --- 3 Tflite layer type --- 19 layer output name--- customed_output_fourth  -- [tidl_tfLiteRtImport_core.cpp, 3090]
[TIDL Import] [PARSER] SUPPORTED: Supported TIDL layer type --- 5 Tflite layer type --- 0 layer output name--- model_5/model_4/model_3/p_re_lu_126/add;model_5/model_4/model_3/p_re_lu_126/Relu;model_5/model_4/model_3/p_re_lu_126/Neg_1;model_5/model_4/model_3/p_re_lu_126/Relu_1;model_5/model_4/model_3/p_re_lu_126/mul1  -- [tidl_tfLiteRtImport_core.cpp, 3090]
[TIDL Import] [PARSER] SUPPORTED: Supported TIDL layer type --- 1 Tflite layer type --- 3 layer output name--- model_5/model_4/model_3/batch_normalization_133/FusedBatchNormV3;model_5/model_4/model_3/depthwise_conv2d_63/depthwise;model_5/model_4/model_3/conv2d_82/Conv2D  -- [tidl_tfLiteRtImport_core.cpp, 3090]
[TIDL Import] [PARSER] SUPPORTED: Supported TIDL layer type --- 5 Tflite layer type --- 18 layer output name--- customed_output_first  -- [tidl_tfLiteRtImport_core.cpp, 3090]

Total Nodes = 7
-------------------------------------------------------------------------------
|          Core           |      No. of Nodes       |   Number of Subgraphs   |
-------------------------------------------------------------------------------
| C7x                     |                       7 |                       1 |
| CPU                     |                       0 |                       x |
-------------------------------------------------------------------------------
============================= [Parsing Completed] =============================

In TIDL_tfliteRtImportInit subgraph_id=490
Layer 0, subgraph id 490, name=customed_output_first
Layer 1, subgraph id 490, name=input_12
In TIDL_tfliteRtImportNode, TIDL Layer type - 1, Tflite builtin code type - 3 
In TIDL_tfliteRtImportNode, TIDL Layer type - 5, Tflite builtin code type - 18 
In TIDL_tfliteRtImportNode, TIDL Layer type - 5, Tflite builtin code type - 18 
In TIDL_tfliteRtImportNode, TIDL Layer type - 3, Tflite builtin code type - 19 
In TIDL_tfliteRtImportNode, TIDL Layer type - 5, Tflite builtin code type - 0 
In TIDL_tfliteRtImportNode, TIDL Layer type - 1, Tflite builtin code type - 3 
In TIDL_tfliteRtImportNode, TIDL Layer type - 5, Tflite builtin code type - 18 
=================== [Optimization for subgraph_490 started] ===================

In TIDL_runtimesOptimizeNet: LayerIndex = 9, dataIndex = 8 
[TIDL Import] [PARSER] WARNING: Batch Norm Layer customed_output_third's coeff cannot be found(or not match) in coeff file, Random bias will be generated only for evaluation usage. Results are all random

[TIDL Import]  ERROR:  - Failed in function: tidl_optimizeNet -- [tidl_import_core.cpp, 2473]
[TIDL Import]  ERROR: Network Optimization failed - Failed in function: TIDL_runtimesOptimizeNet -- [tidl_runtimes_import_common.cpp, 1232]
interpreter created!
batch :  1 new_height :  256 new_width :  256 channel :  3
Segmentation fault (core dumped)

1. How can i solve this error?

2. How can I find the "tidl_import_core.cpp" or "itdl_runtimes_import_common.cpp" or the orignal src file of "tidl_model_import_tflite.so"?

thank you.

8 months ago

0 Asha Bhandarkar 8 months ago

TI__Genius 10170 points

Hi Yujin,

Thank you for providing the debug log in your original error.

In terms of solving the error, would you be able to provide the model you are running so we can try reproducing and investigating further on our end?

In terms of source code, we do not offer most of the code for the TIDL libraries. However, I believe most of those files you can see reference version of in c7x-mma-tidl which is packaged in the RTOS sdk.

Best,

Asha

0 Yujin Park 7 months ago

Prodigy 40 points

Hi asha,

Thanks for your response.

The model is 5305.models.zip.

Actually, I was trying to import the original.tflite but got error.

So, I cut the model and found the smallest one which I started to get error.

The model of above question is cut_model.tflite.

Thanks.

Yujin

0 Yujin Park 7 months ago in reply to Asha Bhandarkar

Prodigy 40 points

Hi asha.

Is there something going on?

I'm still waiting for your response.

Thanks.

Yujin

0 Yujin Park 7 months ago in reply to Asha Bhandarkar

Prodigy 40 points

Hi Asha,

I want to know what is the problem of the model importing.

The model is on above reply.

If you find this, please let me know what the problem is or when you can investigate about error.

Thanks.

Yujin

Processors

Processors forum

TDA4VM: segmentation fault when model import using tidl tools