niSrSSKJr SSKJr SSKJr SSKrSSKrSSKrSSKJ s J r \R"SS5rS SjrS S jrg!\ a SSK r N*f=f) z*Network definitions for the PPO algorithm.)absolute_import)division)print_functionN NetworkOutputz"policy, mean, logstd, value, statec [RRRURS9n[R "UR S5n[R"U[R"U5S[R"U5S[R"[RURR5SSS5/5n[R"S5 UnURHEn [RRR!X[R"R$5nMG [RRR!UU[R&US9n [R("S U RSS[R*U5n [R,"U S [R"U 5S[R"U 5S/S/U RR.S- --5n SSS5 [R"S 5 UnUR0HEn [RRR!X[R"R$5nMG [RRR!USS5S n SSS5 [R2"W S 5n [R2"W S 5n [R2"W S 5n [RR4R7U [R8"U 55n [;XXU5$!,(df  GNC=f!,(df  N=f)aIndependent feed forward networks for policy and value. The policy network outputs the mean action and the log standard deviation is learned as independent parameter vector. Args: config: Configuration object. action_size: Length of the action vector. observations: Sequences of observations. unused_length: Batch of sequence lengths. state: Batch of initial recurrent states. Returns: NetworkOutput tuple. factor绽|=rNpolicyweights_initializerlogstdNNvalue.rmean)tfcontriblayersvariance_scaling_initializerinit_mean_factorrandom_normal_initializer init_logstdreshapeshape functoolsreduceoperatormulas_listvariable_scope policy_layersfully_connectednnrelutanh get_variablefloat32tilendims value_layerscheck_numerics distributionsMultivariateNormalDiagexpr)config action_size observations unused_lengthstatemean_weights_initializerlogstd_initializerflat_observationsxsizerrrr s W/home/james-whalen/.local/lib/python3.13/site-packages/pybullet_envs/agents/networks.pyfeed_forward_gaussianr= s  ZZ..KK  $ $L&33F4F4FNjjhh|Qhh|Qx||#))113AB7<0  "A$$ **   + +ARUUZZ @a% ::   , ,Q-8-/WWAY - [D__Xtzz!"~rzzCU VF WWVJ'hhtnQ'$):;qcTZZEUEUXYEY>ZZ\F# !A## **   + +ARUUZZ @a$ JJ   - -aD 9& AE "  4 ($  VX .&  E7 +% :: # # : :4 P& vVE ::)#""!s0D,M:BM* M'* M8c [RRRURS9n[R "UR S5n[RRRURS5n[R"U[R"U5S[R"U5S[R"[RURR!5SSS5/5n[R""S5 Un URSSHEn [RRR%X[R&R(5n MG [R&R+XyX4[R,5up[RRR%U U[R.US 9n [R0"S U RSS[R,U5n [R2"U S [R"U 5S[R"U 5S/S/U RR4S- --5n SSS5 [R""S 5 Un UR6HEn [RRR%X[R&R(5n MG [RRR%U SS5S n SSS5 [R8"W S5n [R8"W S 5n [R8"W S 5n [RR:R=U [R>"U 55n[AXXU5$!,(df  GNC=f!,(df  N=f)aIndependent recurrent policy and feed forward value networks. The policy network outputs the mean action and the log standard deviation is learned as independent parameter vector. The last policy layer is recurrent and uses a GRU cell. Args: config: Configuration object. action_size: Length of the action vector. observations: Sequences of observations. length: Batch of sequence lengths. state: Batch of initial recurrent states. Returns: NetworkOutput tuple. rr rr r Nr rrrrrr)!rrrrrrrrnn GRUBlockCellr$rrrrr r!r"r#r%r&r' dynamic_rnnr*r(r)r+r,r-r.r/r0r1r)r2r3r4lengthr6r7r8cellr9r:r;rrrr s r<recurrent_gaussianrEPs" ZZ..KK  $ $L&33F4F4FN  $ $V%9%9"%= >$jjhh|Qhh|Qx||#))113AB7<0  "A$$Sb) **   + +ARUUZZ @a*uu  &DHA ::   , ,Q-8-/WWAY - [D__Xtzz!"~rzzCU VF WWVJ'hhtnQ'$):;qcTZZEUEUXYEY>ZZ\F# !A## **   + +ARUUZZ @a$ JJ   - -aD 9& AE "  4 ($  VX .&  E7 +% :: # # : :4 P& vVE ::-#""!s&E O$BO O O")N)__doc__ __future__rrr collectionsrr tensorflow.compat.v1compatv1r Exception tensorflow namedtuplerr=rEr<rQs\1&%##&&8\] -;`1;ms A AA