Haoran You comments

Results 9 comments of


                                            Haoran You

Validation Loss Statistics: min=nan, med=nan, mean=nan, max=nan

![image](https://user-images.githubusercontent.com/18202368/54808365-f7691d80-4cba-11e9-8597-39e6092b92c3.png) The model collapse very quickly with exploded loss. I want to figure out why.

Validation Loss Statistics: min=nan, med=nan, mean=nan, max=nan

It seems nothing to do with initialization methods.

Validation Loss Statistics: min=nan, med=nan, mean=nan, max=nan

I found it! the adam parameter beta_2 should never be 0.01 :)

Validation Loss Statistics: min=nan, med=nan, mean=nan, max=nan

> I found it! the adam parameter beta_2 should never be 0.01 :) Still collapse after few epochs :(

Validation Loss Statistics: min=nan, med=nan, mean=nan, max=nan

![image](https://user-images.githubusercontent.com/18202368/54830231-15527480-4cf3-11e9-8090-a16bb2d2aba9.png) Seems like collapse to one mode after 2/3,000 iterations, is here anybody can give a reason?

I also can't find the file "systolic_array_synth.csv".

Hi, I could find your setting for `Max Precision (bits), Min Precision (bits), M, N` but still do not know the `Area (um^2), Dynamic Power (nW), ...`. Could you kindly...

I also can't find the file "systolic_array_synth.csv".

It seems like unit number 1 for aforementioned parameters meets the results in your `.ipynb` file, thanks.

FPGA implementation

Hi, have you got a chance to sort out the codebase?

FPGA implementation

Ok, no problem, take your time. Looking forward to it.