mosimod

To take advantage of the low neurons number and the run speed of KAN, I want to translate the trained KAN into c and other 'low level' language like "verilog-a"....

Loss suddenly become Nan during training process.

1

train loss: 5.15e-01 | test loss: 5.33e-01 | reg: 1.47e+01 : 44%|▍| 436/1000 [04:19