Michael Voronov comments

Results 18 comments of


                                            Michael Voronov

linear2d_layer_no_bias: add optional argument to disable biases

@jvdp1 yep, I think we should

linear2d_layer_no_bias: add optional argument to disable biases

Perhaps, I should rename it to FeedForward or MLP. Thoughts?

linear2d_layer_no_bias: add optional argument to disable biases

@milancurcic Disregard the naming comment. It belongs in a different PR: #208 I like the maxpool ideas

linear2d_layer_no_bias: add optional argument to disable biases

@milancurcic Should we merge it? I really would like to do it before the dense2d refactoring

Use Neural-Fortran

@certik In this case I would suggest the following: 1. Add parallelism with coarrays 3. Remove manual `sp` so that the project could be compiled to a desired precision 4....

Mv optimizer from the network level to the layer level

@jvdp1 That's actually a great idea. Apart from obvious performance gains, it can simplify code for combined layers. I will arrange everything in similar fashion in my project here: https://github.com/OneAdder/llm.f...

`test_linear2d_layer` fails: `incorrect updated weights`

I have an idea what could have gone wrong but I don't own any PPC hardware to test it. The issue most likely is that PowerPC might yield slightly different...

Implement nf_metrics

I suggest: `accuracy`, `balanced_accuracy`, `precision`, `recall` and `f_beta`