You are here

Line search failed even after resetting Hessian

7 posts / 0 new
Last post
Line search failed even after resetting Hessian
#1

Hello,

I keep getting following messages (the run doesnt fail). Why is this happening? Havent really found an explanation here. Can anyone help?

Thank you

L.

 

core.optimization.LineMinimizer: (3) Inaccurate G! step= 7.45058e-09 Deriv= -12.3859 Finite Diff= 1.0822e+09
core.optimization.LineMinimizer: (3) Inaccurate G! step= 6.06331e-09 Deriv= -385.071 Finite Diff= 1.32981e+09
core.optimization.LineMinimizer: (3) Inaccurate G! step= 3.88794e-07 Deriv= -385.071 Finite Diff= 2.07382e+07
core.optimization.Minimizer: (3) Line search failed even after resetting Hessian; aborting at iter#57

Post Situation: 
Wed, 2017-12-13 16:52
Loki01

Can anyone help please? I have looked for the answer everywhere

Mon, 2017-12-25 08:54
Loki01

Did you get any help?

I'm also having this problem running AbinitioRelax with distance restraints. This thread https://www.rosettacommons.org/node/3388 indicates the error might be due to the input file data.

In my case, I'm assuming the distance restraints are causing the problem. However, I'm getting the error after several rounds of minimization, so I'm not quite sure why it does not fail right away upon reading my cst file? I have 2 types of distance restraints:

# experimental

AtomPair O 198 N 187 BOUNDED 1.9 3.00 1.0 NOE

etc..

# other

AtomPair CB 38 CB 111 GAUSSIANFUNC 5.0 3.0 RAPTX WEIGHT 0.786491

etc...

there is no format error thrown at any point by Rosetta, so I'm guessing the problem is elsewhere?

Thu, 2018-01-04 17:08
iphan

Hi, no I got no help.

I saw the thread but I am not sure what is going on.

Mon, 2018-01-08 10:08
Loki01

Is the anybody who can help me with that? The modelling seem to work but I am not sure if I can trust my decoys.

I am really not sure where else to turn.

 

Tue, 2018-01-16 06:48
Loki01

So I got help from the Baker Lab folks:

You may want to use the option:

-optimization::default_max_cycles 200

to save computing time.  According to Hahnbeom, 200 cycles is good enough rather than the default 2000.  He says you will likely run into the inaccurate G warning less often also.  Regardless, those warnings can be ignored.

Fri, 2018-01-19 15:47
iphan

Thank you very much for your help!

 

Thu, 2018-02-15 13:54
Loki01