loss keeps 4.96 all along

[white2018]

@xuannianz Excellent work.

I have encounter such problems in training process. 2189/2250 [============================>.] - ETA: 27s - loss: 4.96312020-03-09 17:47:41 [INFO] [Thread-6] [geos.py:219] Self-intersection at or near point 507.97936313701194 440.22383226394601 2020-03-09 17:47:41 [INFO] [Thread-6] [geos.py:219] Self-intersection at or near point 639.2710706150342 574.59295807232354 2190/2250 [============================>.] - ETA: 27s - loss: 4.96312020-03-09 17:47:41 [INFO] [Thread-6] [geos.py:219] Self-intersection at or near point 482.31532442427749 637.21180633522147 2196/2250 [============================>.] - ETA: 24s - loss: 4.96312020-03-09 17:47:44 [INFO] [Thread-6] [geos.py:219] Self-intersection at or near point 36.365901518633024 639.57219251336892 2200/2250 [============================>.] - ETA: 22s - loss: 4.9631^C

and the loss keeps 4.9631. any advice to me?

Thanks a lot

[xxxxxxxiao]

Hi! have u fixed the problem? i got the same issue. My loss was able to converge to 1.xx but when i trained it again i got 4.9xx. I believe i did not make any change between those two training procedures.

[its-jd]

Hi, were you able to solve the problem? I am also stuck at this step. Loss always stays at 4.95xxx Any thoughts what might go wrong in the training?

[xxxxxxxiao]

setting a smaller learning rate may work

[nguyenquanghieu2000d]

have you done it yet?? same error

[nguyenquanghieu2000d]

setting a smaller learning rate may work

setting a smaller learning rate may work

i set it to 1e-7 but not work

[duydangg]

i got the same problem but my loss is nan

[duydangg]

i got the same problem but my loss is nan

my datasets caused this problem, i've fixed this but now my val loss is 1.00 at every epoch