Hi,
I was working on implementing the sample maxpool custom-layer provided in the tidl/custom package.
I took the inception v1 tflite model from here: https://github.com/TexasInstruments/edgeai-modelzoo/tree/master/models/vision/classification/imagenet1k/tf1-models
This has maxpool layers with 3x3 stride-2 that fits the example provided.
Below is the config-file I used for import:
modelType = 3
numParamBits = 8
numFeatureBits = 8
inputNetFile = ../../test/testvecs/models/public/tflite/inception_v1_224_quant.tflite
outputNetFile = "../../test/testvecs/config/tidl_models/tflite/tidl_net_tflite_inceptionnet_v1_quant_custom.bin"
outputParamsFile = "../../test/testvecs/config/tidl_models/tflite/tidl_io_tflite_inceptionnet_v1_quant_custom_"
inDataNorm = 1
inMean = 128 128 128
inScale = 0.0078125 0.0078125 0.0078125
resizeWidth = 256
resizeHeight = 256
inWidth = 224
inHeight = 224
inNumChannels = 3
inData = ../../test/testvecs/config/imageNet_sample_val_bg.txt
postProcType = 1
quantizationStyle = 3
perfSimConfig = ../../test/testvecs/config/import/device_config.cfg
enableCustomLayers = 0
I used the 'make it_run command' and included this config as first line in import_all_models.sh
At first I kept enableCustomLayers as 0, the import process was successful and generated the artifacts in output folder.
Later I modified enableCustomLayers to 1 and ran the make command, the import process went successful as before
But when I checked the output svg I couldn't find the Custom layer, I still see the Pooling layer as before.
I'm expecting the top-2 pooling layers to be modified but its not happening.
I checked in the makefile_shared in utils and tidl_custom_import.c was also included.
sdk-version that I'm using is: ti-processor-sdk-rtos-j721e-evm-08_00_00_12
Please let me know if I'm missing any details for custom-layer execution to get this working. Thanks.
