Hi,
When I try to test resnet50 from mlperf on TDA4VM（zenodo.org/.../ssd_mobilenet_v1_coco_2018_01_28.onnx）, which is claimed to support on TDA4VM in <Performance and efficiency benchmarking with TDA4 Edge AI processors>, I meet a problem in importing with tidl_tool as following, how to deal with the issue?
=======================================================================================
jacky.lin@CTOSvr:~/workspace/ti/edgeai-tidl-tools/tidl_tools$ ./tidl_model_import.out resnet50/tidl_import_resnet50.txt
ONNX Model (Proto) File : ./resnet50/resnet50_v1_opset11.onnx
TIDL Network File : ./resnet50/tidl_net_mlperf_resnet50_v1.bin
TIDL IO Info File : ./resnet50/tidl_param_mlperf_resnet50_v1.bin
Current ONNX OpSet Version : 11
Warning : only keepdims = 1 supported for Argmax Layer - forcing it to be 1
ONNX operator MatMul is not suported now.. By passing
ONNX operator ReduceMean is not suported now.. By passing
WARNING: Batch Norm Layer resnet_model/dense/BiasAdd's coeff cannot be found(or not match) in coef file, Random bias will be generated! Only for evaluation usage! Results are all random!
Error: Layer 75, softmax_tensor:softmax_tensor:0 is missing inputs in the network and cannot be topologically sorted
Input 0: resnet_model/dense/BiasAdd:0, dataId=151
~~~~~Running TIDL in PC emulation mode to collect Activations range for each layer~~~~~
Processing config file #0 : /home/jacky.lin/workspace/ti/edgeai-tidl-tools/tidl_tools/resnet50/tidl_import_resnet50.txt.qunat_stats_config.txt
Freeing memory for user provided Net
----------------------- TIDL Process with REF_ONLY FLOW ------------------------
# 0 . ..
~~~~~Running TIDL in PC emulation mode to collect Activations range for each layer~~~~~
Processing config file #0 : /home/jacky.lin/workspace/ti/edgeai-tidl-tools/tidl_tools/resnet50/tidl_import_resnet50.txt.qunat_stats_config.txt
Freeing memory for user provided Net
----------------------- TIDL Process with REF_ONLY FLOW ------------------------
# 0 . ..
***************** Calibration iteration number 0 completed ************************
------------------ Network Compiler Traces -----------------------------
Invalid layer parameters for layer 71, 34
Invalid layer parameters for layer 72, 18
Error : Error Code = <ERR_UNKNOWN>
TIDL ALLOWLISTING LAYER CHECK: TIDL_E_QUANT_STATS_NOT_AVAILABLE] tidl_quant_stats_tool.out fails to collect dynamic range. Please look into quant stats log. This model will get fault on target.
****************************************************
** 0 WARNINGS 1 ERRORS **
****************************************************
=======================================================================================
The TIDL tool is build from latest:
https://github.com/TexasInstruments/edgeai-tidl-tools.git
The import config file (tidl_import_resnet50.txt)：
=======================================================================================
# Default - 0
randParams = 0
# 0: Caffe, 1: TensorFlow, 2: ONNX, Default - 0
modelType = 2
# 0: Fixed quantization By tarininng Framework, 1: Dyanamic quantization by TIDL, Default - 1
quantizationStyle = 1
# quantRoundAdd/100 will be added while rounding to integer, Default - 50
quantRoundAdd = 50
numParamBits = 8
inputNetFile = "./resnet50/resnet50_v1_opset11.onnx"
outputNetFile = "./resnet50/tidl_net_mlperf_resnet50_v1.bin"
outputParamsFile = "./resnet50/tidl_param_mlperf_resnet50_v1.bin"
inData = "../test_data/calibrate_list.txt"
preProcType = 2
sampleInData = "../test_data/airshow.jpg"
tidlStatsTool = "./PC_dsp_test_dl_algo.out"
inWidth = 224
inHeight = 224
inNumChannels = 3
inMean = 123.675 116.28 103.53
inScale = 0.017125 0.017507 0.017429
=======================================================================================
