[参考译文] TDA4VH-Q1：导出模型时发生错误。 TIDL 版本：11.01.04.01

尊敬的 Hongyao：

您可以重新刷新链接吗？ 此链接似乎不起作用。 您是否也从浏览器复制并粘贴了链接、或者您是否单击了通过链接共享？ 应该也提供了访问代码。

就模型导出而言、这是在器件上还是在仿真上？ 请在您共享的文件中分享您的意见和完整日志。  

此致、

Christina Kuruvilla

尊敬的 Adam：

已收到！ 感谢您的评分

此致、

Christina

您好 Hongyao、

我可以重现您的问题：

[TIDL 导入]【解析器】错误：找不到整形图层的输入、 输出名称：/TopK_output_1、输入名称：/TopK_output_1_20_21 [tidl_import_common.cpp, 21649]
[TIDL Import] 错误： — 函数中失败： tidl_optimizeNet --[tidl_import_core.cpp、3183]
[TIDL Import] 致命错误：网络优化失败--[tidl_import_main.cpp、500]
[TIDL 导入]中止

这种重塑层很可能被引入链接第二个图形,这似乎是由一些不受支持的图层引起的。

不支持还原最大值和转置

创建第二个图形时会导致此问题。

我可以通过对您的模型进行以下修改来导入此模型。 我在 sigmod 后删除了一些节点并添加了输出。

总之、这应该是 TIDL 无法正确解析模型的问题、 但是、您可以使用将某些节点进行后处理的解决方法。

此致、

Adam

您好、Adam、

如下图所示、  支持还原最大值和转置。 是否有问题？

github.com/.../supported_ops_rts_versions.md

谢谢。

尊敬的 Hongyao：

这是一个错误,如果转置后发生还原,则运算符的顺序不能正确解析。

我可以使用此简化模型重现您的问题：

e2e.ti.com/.../sigmod_5F00_transpose_5F00_reducemax_5F00_model_5F00_16384_5F00_17_5F00_opset18.onnx

Available execution providers :  ['TIDLExecutionProvider', 'TIDLCompilationProvider', 'CPUExecutionProvider']

Running 1 Models - ['sigmod_transpose_reducemax_opset18']


Running_Model :  sigmod_transpose_reducemax_opset18  


Running shape inference on model ../../../models/public/sigmod_transpose_reducemax_model_16384_17_opset18.onnx 

tidl_tools_path                                 = /home/ht/edgeai/edgeai-tidl-tools/tools/AM69A/tidl_tools 
artifacts_folder                                = ../../../model-artifacts//sigmod_transpose_reducemax_opset18/artifacts 
tidl_tensor_bits                                = 8 
debug_level                                     = 1 
num_tidl_subgraphs                              = 16 
num_tidl_subgraph_max_node                      = 0 
enable_rt_multi_subgraph_support                = 0 
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 = 2 
tidl_calibration_options:bias_calibration_iterations = 5 
mixed_precision_factor = -1.000000 
model_group_id = 0 
power_of_2_quantization                         = 2 
ONNX QDQ Enabled                                = 0 
enable_high_resolution_optimization             = 0 
pre_batchnorm_fold                              = 1 
add_data_convert_ops                            = 3 
output_feature_16bit_names_list                 =  
m_params_16bit_names_list                       =  
m_single_core_layers_names_list                 =  
m_spatial_split_layers_names_list               =  
m_channel_split_layers_names_list               =  
Inference mode                                  = 0 
Number of cores                                 = 1 
reserved_compile_constraints_flag               = 1601 
partial_init_during_compile                     = 0 
packetize_mode                                  = 0 
enable_tfr_optimization                         = 0 
ti_internal_reserved_1                          = 

========================= [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          |              11_01_06_00             |
-------------------------------------------------------------------------------
|         C7x Firmware Version         |              11_00_08_00             |
-------------------------------------------------------------------------------
|            Runtime Version           |                1.15.0                |
-------------------------------------------------------------------------------
|          Model Opset Version         |                  18                  |
-------------------------------------------------------------------------------

============================== [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: Layers type supported by TIDL --- layer type - Sigmoid,  Node name - /hidden/Sigmoid -- [tidl_onnxRtImport_core.cpp, 652]
[TIDL Import] [PARSER] UNSUPPORTED: Layer 1 - op type ReduceMax, Unknown input dimension, not supported by TIDL -- [tidl_onnxRtImport_core.cpp, 582]
[TIDL Import] [PARSER] UNSUPPORTED: Layer 2 - op type Transpose, Unknown input dimension, not supported by TIDL -- [tidl_onnxRtImport_core.cpp, 582]

------------------------- Subgraph Information Summary -------------------------
-------------------------------------------------------------------------------
|          Core           |      No. of Nodes       |   Number of Subgraphs   |
-------------------------------------------------------------------------------
| C7x                     |                       1 |                       1 |
| CPU                     |                       2 |                       x |
-------------------------------------------------------------------------------
--------------------------------------------------------------------------------------------------------
|   Node    |  Node Name |                                   Reason                                    |
--------------------------------------------------------------------------------------------------------
| ReduceMax | /ReduceMax | Layer 1 - op type ReduceMax, Unknown input dimension, not supported by TIDL |
| Transpose | Transpose  | Layer 2 - op type Transpose, Unknown input dimension, not supported by TIDL |
--------------------------------------------------------------------------------------------------------
Running Runtimes GraphViz - /home/ht/edgeai/edgeai-tidl-tools/tools/AM69A/tidl_tools/tidl_graphVisualiser_runtimes.out ../../../model-artifacts//sigmod_transpose_reducemax_opset18/artifacts/allowedNode.txt ../../../model-artifacts//sigmod_transpose_reducemax_opset18/artifacts/tempDir/graphvizInfo.txt ../../../model-artifacts//sigmod_transpose_reducemax_opset18/artifacts/tempDir/runtimes_visualization.svg 
============================= [Parsing Completed] =============================

TIDL_createStateImportFunc Started:
Compute on node : TIDLExecutionProvider_TIDL_0_0
  0,         Sigmoid, 1, 1, input, /hidden/Sigmoid_output_0

Input tensor name -  input 
Output tensor name - /hidden/Sigmoid_output_0 
Process Process-1:
Traceback (most recent call last):
  File "/home/ht/.pyenv/versions/3.10.18/lib/python3.10/multiprocessing/process.py", line 314, in _bootstrap
    self.run()
  File "/home/ht/.pyenv/versions/3.10.18/lib/python3.10/multiprocessing/process.py", line 108, in run
    self._target(*self._args, **self._kwargs)
  File "/home/ht/edgeai/edgeai-tidl-tools/examples/osrt_python/ort/onnxrt_ep.py", line 385, in run_model
    height = input_details[0].shape[2]
IndexError: list index out of range
************ in TIDL_subgraphRtDelete ************ 
^CTraceback (most recent call last):
  File "/home/ht/edgeai/edgeai-tidl-tools/examples/osrt_python/ort/onnxrt_ep.py", line 579, in <module>
    nthreads = join_one(nthreads)
  File "/home/ht/edgeai/edgeai-tidl-tools/examples/osrt_python/ort/onnxrt_ep.py", line 546, in join_one
    sem.acquire()
KeyboardInterrupt

此致、

Adam