h" ZSSKrSSKJr SSKrSSKrSSKJr SSK J r S\RS\ 4SjrSS\RS\ S \ S\ 4S jjrS \R"S\RS\R"4S jrSS \\R&\\\R&44S\RS\ S\\R&\\\R&444SjjrS\RS\\\S4\\\\S4444SjrS\RS\4SjrS\RS\4SjrS\RSS4Sjrg)N)Union)spaces) functionalobservation_spacereturnc[R"UR5R5nUS:Xa[R "S5 US:H$)av Check if an image observation space (see ``is_image_space``) is channels-first (CxHxW, True) or channels-last (HxWxC, False). Use a heuristic that channel dimension is the smallest of the three. If second dimension is smallest, raise an exception (no support). :param observation_space: :return: True if observation space is channels-first image, False if channels-last. zXTreating image space as channels-last, while second dimension was smallest of the three.r)npargminshapeitemwarningswarn)rsmallest_dimensions `/home/james-whalen/.local/lib/python3.13/site-packages/stable_baselines3/common/preprocessing.pyis_image_space_channels_firstr sC#4#:#:;@@BQ pq  ""check_channelsnormalized_imagecU(+=p4[U[R5(a[UR5S:XaU(aUR [ R:wag[ R"URS:g5=(d# [ R"URS:g5nU(aU(agU(dg[U5(aURSnUS;$URSnUS;$g)a Check if a observation space has the shape, limits and dtype of a valid image. The check is conservative, so that it returns False if there is a doubt. Valid images: RGB, RGBD, GrayScale with values in [0, 255] :param observation_space: :param check_channels: Whether to do or not the check for the number of channels. e.g., with frame-stacking, the observation space may have more channels than expected. :param normalized_image: Whether to assume that the image is already normalized or not (this disables dtype and bounds checks): when True, it only checks that the space is a Box and has 3 dimensions. Otherwise, it checks that it has expected dtype (uint8) and bounds (values in [0, 255]). :return: FrT)r r) isinstancerBoxlenr dtyper uint8anylowhighr)rrr check_dtype check_boundsincorrect_bounds n_channelss ris_image_spacer's*&6!55K#VZZ00S9J9P9P5QUV5V ,22bhh>66"3"7"71"<=fHYH^H^beHeAf , (): ; ;*003JY&&+004JY&& r observationc4SSKJn [U5(aURUR:XdgURSSUR:XdJUR U5nURUR:XdURSSUR:XaUnU$)z Handle the different cases for images as PyTorch use channel first format. :param observation: :param observation_space: :return: channel first observation if observation is an image r)VecTransposeImager N) stable_baselines3.common.vec_envr*r'r transpose_image)r(rr* transpose_obss rmaybe_transposer.HsC'((!!%6%<%<< @Q@QRSRT@UYjYpYp@p-==kJM""&7&=&==ATATUVUWAX\m\s\sAs+ robsnormalize_imagesc [U[R5(aT[U[5(dS[ U535e0nUR 5HupE[ XQUUS9X4'M U$[U[R5(dS[ U535e[U[R5(a:U(a#[U5(aUR5S- $UR5$[U[R5(aD[R"UR5[!UR"5S9R5$[U[R$5(a[R&"[)[R*"UR5SSS95VVs/sHLupg[R"UR5[!UR,U5S9R5PMN snnSS9R/UR0S [3UR,55$[U[R45(aUR5$[7S U35es snnf) aP Preprocess observation to be to a neural network. For images, it normalizes the values by dividing them by 255 (to have values in [0, 1]) For discrete observations, it create a one hot vector. :param obs: Observation :param observation_space: :param normalize_images: Whether to normalize images or not (True by default) :return: zExpected dict, got )r0z"Expecting a torch Tensor, but got go@) num_classesr )dimrrz"Preprocessing not implemented for )rrDictdicttypeitemspreprocess_obsthTensorrr'floatDiscreteFone_hotlongintn MultiDiscretecat enumeratesplitnvecviewr sum MultiBinaryNotImplementedError)r/rr0preprocessed_obskey_obsidxobs_s rr8r8\s #V[[11#t$$G(;DI;&GG$IC$2439Obr$s  !% c299 % %W)KDQTI;'WW %#VZZ00 /@ A A99;& &yy{ %v 7 7yy5F5H5H1IJPPRR %v';'; < <vv"+288CHHJq+I!J !JIC $))+37H7M7Mc7R3STZZ\!J   $syy|S!2!7!78 9  : %v'9'9 : :yy{!$FGXFY"Z[[ sAJ .c0[U[R5(a UR$[U[R5(ag[U[R 5(a[ UR54$[U[R5(a UR$[U[R5(a9URR5VVs0sHupU[U5_M snn$[US35es snnf)z` Get the shape of the observation (useful for the buffers). :param observation_space: :return: )r z# observation space is not supported) rrrr r<rBrrFrIr4r7 get_obs_shaperJ)rrLsubspaces rrQrQs#VZZ00 &&& %v 7 7 %v';'; < <%**+-- %v'9'9 : : &&& %v{{ 3 3DUD\D\DbDbDdeDd#]8,,Ddee"%6$77Z"[\\fs(Dc[U[R5(a[UR5$[R R U5$)z Get the dimension of the observation space when flattened. It does not apply to image observation space. Used by the ``FlattenExtractor`` to compute the input shape. :param observation_space: :return: )rrrBrHrFutilsflatdim)rs rget_flattened_obs_dimrVsB#V%9%9::$))**||##$566r action_spacec&[U[R5(a)[[R "UR 55$[U[R5(ag[U[R5(a[UR5$[U[R5(aI[UR[5(dSURS35e[UR5$[US35e)zG Get the dimension of the action space. :param action_space: :return: r zMulti-dimensional MultiBinary(z<) action space is not supported. You can flatten it instead.z action space is not supported)rrrr@r prodr r<rBrrFrIrArJ)rWs rget_action_dimrZs, ++277<--.// L&// 2 2 L&"6"6 7 7<$$%% L&"4"4 5 5 NNC   y +LNN+;;w x y <>>""!\N2P"QRRr obs_spacect[U[R[R45(a[U[R5(aURR 5O URnUH=n[U[R[R45(dM4[ S5e gg)z Make sure the observation space does not have nested spaces (Dicts/Tuples inside Dicts/Tuples). If so, raise an Exception informing that there is no support for this. :param obs_space: an observation space zWNested observation spaces are not supported (Tuple/Dict space inside Tuple/Dict space).N)rrr4TuplevaluesrJ)r[ sub_spaces sub_spaces rcheck_for_nested_spacesras)fkk6<<8992rms$#VZZ#D#&!"*||*** *Z QSQ[Q[."0\ ryy$sBII~.. /0\||0\0\ 299d3 >* *+ 0\f]||] 5c?DeCHo!56 67]67V\\7c7&SS#S2 v||  r