tensorflow 0.11.0, amp v0.6

[Simon Muller]

I am running the stable release v0.6 using tf 0.11.0. I pass all the tests, however, when running the tests, tensorflow give a lot of verbose output. Is the output something that one should worry? Any help would be very much appreciated.

It's of the following type:

... Level 1:tensorflow:Gradient for 'MatMul_3' Level 1:tensorflow: in --> gradients_3/add_7_grad/tuple/control_dependency:0 Level 1:tensorflow: out --> gradients_3/MatMul_3_grad/tuple/control_dependency:0, gradients_3/MatMul_3_grad/tuple/control_dependency_1:0 Level 1:tensorflow:Gradient for 'af61d8ba_Pt_bfc0/read' Level 1:tensorflow: in --> gradients_3/add_7_grad/tuple/control_dependency_1:0 Level 1:tensorflow: out --> gradients_3/add_7_grad/tuple/control_dependency_1:0 Level 2:tensorflow: _AggregatedGrads 2 x (20, 3) using add_n Level 1:tensorflow:Gradient for 'af61d8ba_Cu_Wfc0/read' Level 1:tensorflow: in --> gradients_3/AddN_12:0 Level 1:tensorflow: out --> gradients_3/AddN_12:0 Level 2:tensorflow: _AggregatedGrads 2 x (20, 3) using add_n Level 1:tensorflow:Gradient for 'af61d8ba_Pt_Wfc0/read' Level 1:tensorflow: in --> gradients_3/AddN_13:0 Level 1:tensorflow: out --> gradients_3/AddN_13:0

[Muammar El Khatib]

I think it is normal. Is tests/test_gaussian_tflow.py giving you results that make sense?

[andrew_peterson]

Closing this for now. Please post to amp-users if further questions.