nivSrSSKJr SSKJr SSKJr SSKJs Jr "SS\ 5r g!\ a SSK rNf=f)Memory that stores episodes.)absolute_import)division)print_functionNcP\rSrSrSrSrS SjrS SjrS SjrS Sjr S S jr S r g) EpisodeMemoryrc X lX0l[R"U5nXPl[R "[R "U[R5S5UlUVs/sHUn[R "[R "X#/URR5-UR5S5PMW snUl SSS5 gs snf!,(df  g=f)aCreate a memory that stores episodes. Each transition tuple consists of quantities specified by the template. These quantities would typically be be observartions, actions, rewards, and done indicators. Args: template: List of tensors to derive shapes and dtypes of each transition. capacity: Number of episodes, or rows, hold by the memory. max_length: Allocated sequence length for the episodes. scope: Variable scope to use for internal variables. FN) _capacity _max_lengthtfvariable_scope_scopeVariablezerosint32_lengthshapeas_listdtype_buffers)selftemplatecapacity max_lengthscope var_scopeelems Y/home/james-whalen/.local/lib/python3.13/site-packages/pybullet_envs/agents/ppo/memory.py__init__EpisodeMemory.__init__sN!  5 !Yk[[(BHH!=uEdld ++bhh5 8J8J8LLdjjY[` adm " ! " !sA C-AC CC C-NcUc [R"UR5OUn[R"URU5$)zTensor holding the current length of episodes. Args: rows: Episodes to select length from, defaults to all. Returns: Batch tensor of sequence lengths. )r ranger gatherrrrowss rlengthEpisodeMemory.length4s1(,|288DNN #D 99T\\4 ((c "Uc [R"UR5OUnURRS:Xde[R "X RSS9n[R "U/5 [R "[R"URU5URSS9nSSS5 /n[R "W/5 [URU5Hdupg[R"URU5n[R"X(/S5n UR"XiU55 Mf SSS5 [R "U5 [R"[R "X R[R"S9S5n URR%U 5sSSS5 $!,(df  GN,=f!,(df  N=f!,(df  g=f)zAppend a batch of transitions to rows of the memory. Args: transitions: Tuple of transition quantities with batch dimension. rows: Episodes to append to, defaults to all. Returns: Operation. Ncapacity exceededmessagemax length exceeded)rr)r r#r rndims assert_lesscontrol_dependenciesr$rr ziprstackappendscatter_nd_update reduce_sumone_hotr assign_add) r transitionsr&assert_capacityassert_max_length append_opsbuffer_elementstimestepindices episode_masks rr5EpisodeMemory.append@sm(,|288DNN #D ::  q  nnT>>CVWO /!2 3..4<<)F)-)9)91FH 4J "3!4 5"4==+> '99T\\40((D+Q/"..wJK ? 6  ,]]2::dNN"((#SUVWl \\ $ $\ 2 - , 4 3 6 5 - ,s'8AGA?G/6AH G,/ G= HcUc [R"UR5OUnURRS:Xde[R "X0RSS9n[R "U/5 [R"UURSS9nSSS5 /n[R "W/5 [URU5H-upx[R"XsU5n URU 5 M/ SSS5 [R "U5 [R"URX25sSSS5 $!,(df  N=f!,(df  N`=f!,(df  g=f)zReplace full episodes. Args: episodes: Tuple of transition quantities with batch and time dimensions. length: Batch of sequence lengths. rows: Episodes to replace, defaults to all. Returns: Operation. Nr+r,r-r/)r r#r rr0r1r2assert_less_equalr r3rscatter_updater5r) repisodesr'r&r;r< replace_opsr>r? replace_ops rreplaceEpisodeMemory.replace[s(,|288DNN #D ::  q  nnT>>CVWO /!2 3..v/3/?/?7LN 4K "3!4 5"4==(; '&&wh? :& < 6  -   t||T : . - 4 3 6 5 . -s%8!E :AE !E- E E*- E;c6Uc [R"UR5OUnURRS:XdeUR Vs/sHn[R "X!5PM nn[R "URU5nX44$s snf)a@Access a batch of episodes from the memory. Padding elements after the length of each episode are unspecified and might contain old data. Args: rows: Episodes to select, defaults to all. Returns: Tuple containing a tuple of transition quantiries with batch and time dimensions, and a batch of sequence lengths. r+)r r#r rr0rr$r)rr&r>episoder's rdataEpisodeMemory.datausw(,|288DNN #D ::  q  7;}}E}Gryy'}GE YYt||T *F ?Fs BcUc [R"UR5OUnURRS:Xde[R "UR U[R"U55$)zReset episodes in the memory. Internally, this only sets their lengths to zero. The memory entries will be overridden by future calls to append() or replace(). Args: rows: Episodes to clear, defaults to all. Returns: Operation. r+)r r#r rr0rFr zeros_liker%s rclearEpisodeMemory.clearsV(,|288DNN #D ::  q   T\\4t1D EEr))rr rr r)N) __name__ __module__ __qualname____firstlineno____doc__r r'r5rJrNrR__static_attributes__r)rrrs$$. )36;4&Fr)r) rX __future__rrrtensorflow.compat.v1compatv1r Exception tensorflowobjectrrZr)rrbsB#&%## |FF|F s + 88