Fix bug in render_network

[feynmanliang]

L121 makes it so that index is never 0

---

This change is 

[msultan]

ahh good catch with the index, I think it ended up not too much of an issue because I switched towards called in input features layer (cell number [65] ) "l0" instead of "f0".

[msultan]

I am guessing you have 1 feature in the input so this was causing an issue?

[feynmanliang]

I encountered this on the 2-alanine example, so there's 4 (2 sin and 2 cos).

[msultan]

hmm, weird, i will investigate this further then.

[msultan]

This code gives the correct behavior without in both instances if lp is a list but it is possible that pytorch has changed how it returns the bias when tolist() is called.

In [12]: lp=[1]

In [13]: ','.join(map(str,lp))
Out[13]: '1'

In [14]: lp=[1,2,3,4]

In [15]: ','.join(map(str,lp))
Out[15]: '1,2,3,4'

[msultan]

basically, we only need a str that can be streamed to text file for plumed

[feynmanliang]

Yes, the problem I was running in to was that tolist was returning a scalar, i.e .','.join(map(str, 1)), which results in a TypeError since 1 is not iterable

[msultan]

if you are done with the PR, i can merge the changes. Thanks for bringing these in, I really appreciate it.

[feynmanliang]

Sure go ahead and merge, I'll send more as I work.

FYI I'm using the vde package under msmbuilder

[msultan]

this should be lp.bias[i] not lb.bias[i]

[feynmanliang]

Sorry about that, fixed!