{
  "ticketNumber" : "693319",
  "reporterName" : "Navin Prashath",
  "rankPoints" : "760",
  "resolutionStatus" : "",
  "ticketName" : "TDA2: Error when importing Lenet model",
  "rankName" : "Intellectual",
  "replies" : "",
  "views" : "",
  "queryText" : "Part Number: TDA2 Hi, I'm importing the Caffe LeNet model and the tool crashes when processing Fully Connected Layer 1. The inputs to the fc1 layer is 9216 (64x12x12 Flattened) and num of outputs are 1024. Here is the import file. # Default - 0 randParams = 0 # 0: Caffe, 1: TensorFlow, Default - 0 modelType = 0 # 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 = 25 numParamBits = 8 # 0 : 8bit Unsigned, 1 : 8bit Signed Default - 1 inElementType = 0 inputNetFile = \"..\\..\\test\\testvecs\\config\\pp_lenet_model\\lenet.prototxt\" inputParamsFile = \"..\\..\\test\\testvecs\\config\\pp_lenet_model\\lenet.caffemodel\" outputNetFile = \"..\\..\\test\\testvecs\\config\\tidl_models\\NET_OD.BIN\" outputParamsFile = \"..\\..\\test\\testvecs\\config\\tidl_models\\PRM_OD.BIN\" rawSampleInData = 1 preProcType = 4 sampleInData = \"..\\..\\test\\testvecs\\input\\lenet_48_48.y\" tidlStatsTool = \"..\\quantStatsTool\\eve_test_dl_algo.out.exe\" Regards, Navinprashath.R.R",
  "imageList" : null,
  "partNumber" : "NA",
  "allResponseList" : [ {
    "contentId" : "",
    "userName" : "Praveen Eppa1",
    "rankPoints" : "17580",
    "rankName" : "TI__Genius",
    "date" : "",
    "userId" : "/members/6019814",
    "content" : "Hi, Can please try with caffe jacinto model first, may be cifar10. Thanks, Praveen",
    "imageList" : null
  }, {
    "contentId" : "",
    "userName" : "Navin Prashath",
    "rankPoints" : "760",
    "rankName" : "Intellectual",
    "date" : "",
    "userId" : "/members/4510103",
    "content" : "Hi Praveen, In one of the documents, I found that the maximum number of input or output nodes supported for FC layer in TIDL is 4096. So I made a workaround by slicing input nodes(Instead of one FC, now I have 3 FC layers). Regards, Navinprashath.R.R",
    "imageList" : null
  } ],
  "tags" : [ ],
  "fourmType" : "processors-forum"
}