Tool/software:
Hii
I am trying to convert an ONNX model using the TIDL model import tool from the TI Processor SDK RTOS (TDA4VM platform).
During the conversion process, it is asking for the file imageNet_sample_val.txt, but I am unable to find this file anywhere in the SDK or example directories.
Where can I get the imageNet_sample_val.txt file, or is there an alternative/format to create it manually for running the import tool?
Below i have attached a file regarding the error .
test@pcz-mmalab-5:/u/new_sdk/RTOS/ti-processor-sdk-rtos-j721e-evm-11_00_00_06/c7x-mma-tidl/ti_dl/utils/tidlModelImport/out$ ./tidl_model_import.out /u/new_sdk/RTOS/ti-processor-sdk-rtos-j721e-evm-11_00_00_06/c7x-mma-tidl/ti_dl/test/testvecs/config/import/public/pnet/tidl_import_onnx.txt ========================= [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_00_06_00 | ------------------------------------------------------------------------------- | C7x Firmware Version | 11_00_00_00 | ------------------------------------------------------------------------------- ONNX model (Proto) file : /u/new_sdk/RTOS/ti-processor-sdk-rtos-j721e-evm-11_00_00_06/c7x-mma-tidl/ti_dl/test/testvecs/models/public/pnet/pnet.onnx TIDL network file : /u/new_sdk/RTOS/ti-processor-sdk-rtos-j721e-evm-11_00_00_06/c7x-mma-tidl/ti_dl/test/testvecs/config/import/public/pnet/tidl_net_pnet.bin TIDL IO info file : /u/new_sdk/RTOS/ti-processor-sdk-rtos-j721e-evm-11_00_00_06/c7x-mma-tidl/ti_dl/test/testvecs/config/import/public/pnet/tidl_io_pnet_ Current ONNX OpSet version : 13 [TIDL Import] WARNING: Cannot read dims for variable input of Add/Mul operator, shape inference should be done on the model. Assuming 1D const input is channel broadcast, operator maps to BatchNorm layer [TIDL Import] WARNING: Cannot read dims for variable input of Add/Mul operator, shape inference should be done on the model. Assuming 1D const input is channel broadcast, operator maps to BatchNorm layer [TIDL Import] WARNING: Cannot read dims for variable input of Add/Mul operator, shape inference should be done on the model. Assuming 1D const input is channel broadcast, operator maps to BatchNorm layer ============================ [Optimization started] ============================ [TIDL Import] [PARSER] WARNING: Cannot read dims for variable input of Add/Mul operator, shape inference should be done on the model. Assuming 1D const input is channel broadcast, operator maps to BatchNorm layer [TIDL Import] [PARSER] WARNING: Cannot read dims for variable input of Add/Mul operator, shape inference should be done on the model. Assuming 1D const input is channel broadcast, operator maps to BatchNorm layer [TIDL Import] [PARSER] WARNING: Cannot read dims for variable input of Add/Mul operator, shape inference should be done on the model. Assuming 1D const input is channel broadcast, operator maps to BatchNorm layer [TIDL Import] [PARSER] WARNING: Conv Layer p_net/conv1/BiasAdd's coeff cannot be found(or not match) in coeff file, Random coeff will be generated! Only for evaluation usage! Results are all random! ----------------------------- Optimization Summary ----------------------------- -------------------------------------------------------------------------------- | Layer | Nodes before optimization | Nodes after optimization | -------------------------------------------------------------------------------- | TIDL_BatchNormLayer | 0 | 13 | | TIDL_NegLayer | 3 | 0 | | TIDL_ReLULayer | 6 | 0 | | TIDL_EltWiseLayer | 6 | 3 | | TIDL_ConvolutionLayer | 5 | 5 | | TIDL_TransposeLayer | 3 | 3 | | TIDL_PoolingLayer | 1 | 1 | -------------------------------------------------------------------------------- Total nodes in subgraph: 28 =========================== [Optimization completed] =========================== [TIDL Import] ERROR: Could not open /u/new_sdk/RTOS/ti-processor-sdk-rtos-j721e-evm-11_00_00_06/c7x-mma-tidl/ti_dl/test/testvecs/config/imageNet_sample_val.txt file for reading, check if the file exists and is readable -- [tidl_import_core.cpp, 573] Couldn't open inData file: /u/new_sdk/RTOS/ti-processor-sdk-rtos-j721e-evm-11_00_00_06/c7x-mma-tidl/ti_dl/test/testvecs/config/imageNet_sample_val.txt , Skipping Range Collection for Quantization ------------------ Fixed-point Calibration Iteration [1 / 1]: ------------------ Couldn't open inData file: /u/new_sdk/RTOS/ti-processor-sdk-rtos-j721e-evm-11_00_00_06/c7x-mma-tidl/ti_dl/test/testvecs/config/imageNet_sample_val.txt , Skipping Range Collection for Quantization ==================== [Quantization & Calibration Completed] ==================== [TIDL Import] [PARSER] UNSUPPORTED: All the tensor dimensions has to be greater then zero. For tensor p_net/pool1/MaxPool:0, id 9 - Dim 4 is 0 -- [tidl_import_common_model_check.cpp, 2324] [TIDL Import] WARNING: Couldn't open graphVizTool file: ../../utils/tidlModelGraphviz/out/tidl_graphVisualiser.out . Skipping Graph Visualization. [TIDL Import] [PARSER] WARNING: ******************************************************************** * Network compiler returned with error or didn't executed * * This model can only be used on PC/Host emulation mode * * It is not expected to work on target/EVM * ******************************************************************** ======================== Subgraph Compiled Successfully ========================
