Milan Le
Milan Le
So it turns out, the solution was quite easy. Since it is fine-tuning process the learning rate of ConvNet (VGG16, ZFNet, LeNet) must be set to zero. Just put down...
Hi @JulianoLagana, did you managed to solve the problem? I did try to clip all the `np.exp` expressions to some value though still failing due to signal 8: SIGFPE (floating...
Hi @JulianoLagana . I just figured out that the weights of `conv5_3` and lower (conv5_{2,1}, conv4_{1,2,3}, etc) contains NaNs. So the reason might be in bad initialization/loading of the network...
Check #53 for solution
@glaforge I want to contribute
@glaforge sorry for being idle. I could not implement the feature
@langchain4j I will work on that next week
@langchain4j, I drafted implementation and will welcome any comments and suggestions
/test all
cc @yuzisun