Closed giobus75 closed 3 years ago
I not sure if I have understood the problem. Can you elaborate on it a bit more? Btw, the last release is v0.8a.
Hi, I mean that the cout at row 794 in https://github.com/deephealthproject/eddl/blob/v0.8a/src/tensor/tensor_core.cpp (and not present in the v0.7.1) is probably causing some unwanted print of dots at the screen. Specifically, this happens during the evaluation step as I call the forward function (row 188 in the code of my application https://github.com/deephealthproject/promort_pipeline/blob/master/python/promort_cassandra.py) In the attached image, the output of my program:
By the way, you can get a similar behavior running the example mnist_loss (or mnist_rnn_func.cpp or mnist_mlp_func.cpp). This is a sample of mnist_loss output:
You don't get the same behavior by using the other mnist examples that do not call the forward function. This issue is present both using the cpu and the gpu. I hope it is useful :-). Giovanni
okay, understood. We'll try to use the same printing as in the other examples.
Ok new Release we will remove this dots
Hi, using the forward function during the evaluation, I got several dots as video output. Dots are printed after some batches are processed (in my application every 205 batches).
Digging a bit in the code I found out that it is probably due to the function
void Tensor::select(Tensor *A, Tensor *B, vector<int> sind, int ini, int end, bool mask_zeros)
. Maybe thecout<<".....";
at row 794 could be the issue.I did not have this problem with the tagged version 0.7.1.
Thank you.