Part Number: TDA4VM
Dear TI team,
I want to multiply an input tensor by a given static tensor. Converting Element-wise operation to TIDL_EltWiseLayer fails.
Creating a simple onnx model (v.1.3.0, opset 9 ) with Mul operation with 3 different test cases.
Testing with TIDL v02_00_00_07 and 01_04_00_08
Case A. Multiply an input by a tensor given as initializer: Mul_initializer_130.onnx
Fails: with output
****************************************************
** All the Input Tensor Dimensions has to be greater then Zero
** DIM Error - For Tensor 1, Dim 1 is -1
****************************************************
Case B. Multiply Two inputs: Mul_2inputs_130.onnx
Fails: with output
****************************************************
** All the Input Tensor Dimensions has to be greater then Zero
** DIM Error - For Tensor 1, Dim 1 is -1
****************************************************
Case C: Multiply input by itself: Mul_self_130.onnx
OK:
----------------------- TIDL Process with REF_ONLY FLOW ------------------------
# 0 . .. T 0.01 .... ..... ... .... .....
------------------ Network Compiler Traces -----------------------------
successful Memory allocation
****************************************************
** ALL MODEL CHECK PASSED **
****************************************************
* Same result with Add Operation
* Same results with ONNX opset 11
* Inference configuration:
inWidth = 32
inHeight = 16
inNumChannels = 1
inDataFormat = 1
inFileFormat = 0
inData = ".../sample1.png"
- Why Mul is only working for self-multiplication? Could you please give a hint for converting Case A and B ?
Best
Gildardo L.
