do you have somewhere Sparse fused lasso code for logistric regression about liboptpy

[amkatrutsa]

No, I did not implement it, but you can implement the objective function and its (sub)gradient and then use any of the appropriate methods implemented in this library like subgradient descent or proximal-type methods.
Examples of using this library for particular optimization problems you can find here.

[Sandy4321]

are sure it is so easy
seen this link
https://stats.stackexchange.com/questions/364323/solving-linear-regression-with-fused-lasso-regularization-by-mle
After some research I have found out that there is no closed form solution for this problem (and therefore no MLE in particular). A (fast) solution for this problem has been proposed via solution paths, the algorithm can be found here.
??

[amkatrutsa]

Well, probably there exists specific methods for this particular problem, but you also can use any general-purpose solvers if the efficiency is not crucial.