Jan Jongboom

Here's the overview of active memory on the heap (generated via [mbed-find-dangling-ptrs](https://github.com/janjongboom/mbed-find-dangling-pointers/blob/master/README.md)): ``` Extracting symbols from BUILD/DISCO_L475VG_IOT01A/GCC_ARM-DEBUG/firmware.elf Extracting symbols OK Free for untracked pointer 0x10004900 Free for untracked pointer 0x100048e8...

@neil-tan My feeling for this API would be that uTensor should not handle input / output memory. The application passes in the input layer, with memory owned by the application;...

@dboyliao What do you mean? I'm not using any explicit softmax as far as I know. Just simple MLP 40x20x10x4.

@dboyliao ``` from keras.models import Sequential from keras.layers import Dense, InputLayer from keras import regularizers r = regularizers.l1(0.0001) model = Sequential() model.add(InputLayer(input_shape=(X_train.shape[1], ), name='x_input')) model.add(Dense(20, activation='relu', activity_regularizer=r )) model.add(Dense(10, activation='relu',...

@dboyliao @mbartling Does landing #171 mean this issue should be fixed?

@dboyliao @neil-tan I've tested this against the `fix-softmax` branch, updated my graph to (this might be incorrect, I'm not sure?): ```cpp { ctx.add(new RamTensor(), "y_pred/Softmax:0"); ctx.push(new SoftmaxOp(), { "y_pred/BiasAdd:0" },...

@dboyliao @neil-tan Could you confirm that I'm using the API correct here? https://github.com/janjongboom/utensor-misclassifies/blob/e0bfb95df3cdd97a5a665e055e7f656e9797ad64/fw/main.cpp#L30 If the softmax issue is now fixed, I might be using the API wrongly?

I have looked into the fix @dboyliao provided for SoftmaxOp (which was indeed broken). But this is an issue at a deeper end as the values are already incorrect when...

I have put a (I think also helpful for other things) tool here that compares output between uTensor and TF: https://github.com/janjongboom/utensor-fuzzer It currently finds 383 cases in the associated test...

@dboyliao I've added the simplest case that I can think of too here: https://github.com/janjongboom/utensor-fuzzer/tree/super-simple-usecase. Just a two-layer NN (one input layer, then softmax layer) and two examples (one is OK,...