§ O’j<ãóŠ—ddlZddlmZddlmZddlmZddlmZddlm Z ddgZ Gd „de¦«Z Gd „de ¦«Z dS) éN)ÚTensor)Ú constraints)Ú Categorical)Ú Distribution)Ú_sizeÚOneHotCategoricalÚ OneHotCategoricalStraightThroughc ó–‡—eZdZdZejejdœZejZ dZ dde dzde dzde dzddfˆfd „ Z dˆfd „ Zd „Zede fd „¦«Zede fd „¦«Zede fd„¦«Zede fd„¦«Zede fd„¦«Zede fd„¦«Zedejfd„¦«Zej¦«fd„Zd„Zd„Zdd„ZˆxZS)raµ Creates a one-hot categorical distribution parameterized by :attr:`probs` or :attr:`logits`. Samples are one-hot coded vectors of size ``probs.size(-1)``. .. note:: The `probs` argument must be non-negative, finite and have a non-zero sum, and it will be normalized to sum to 1 along the last dimension. :attr:`probs` will return this normalized value. The `logits` argument will be interpreted as unnormalized log probabilities and can therefore be any real number. It will likewise be normalized so that the resulting probabilities sum to 1 along the last dimension. :attr:`logits` will return this normalized value. See also: :func:`torch.distributions.Categorical` for specifications of :attr:`probs` and :attr:`logits`. Example:: >>> # xdoctest: +IGNORE_WANT("non-deterministic") >>> m = OneHotCategorical(torch.tensor([ 0.25, 0.25, 0.25, 0.25 ])) >>> m.sample() # equal probability of 0, 1, 2, 3 tensor([ 0., 0., 0., 1.]) Args: probs (Tensor): event probabilities logits (Tensor): event log probabilities (unnormalized) )ÚprobsÚlogitsTNr r Ú validate_argsÚreturnc󺕗t||¦«|_|jj}|jjdd…}t ¦« |||¬¦«dS)Néÿÿÿÿ©r )rÚ _categoricalÚ batch_shapeÚ param_shapeÚsuperÚ__init__)Úselfr r r rÚ event_shapeÚ __class__s €úUC:\Projects\DSCons\.venv\Lib\site-packages\torch\distributions\one_hot_categorical.pyrzOneHotCategorical.__init__1s[ø€õ (¨¨vÑ6Ô6ˆÔØÔ'Ô3ˆ ØÔ'Ô3°B°C°CÔ8ˆ å ‰Œ×Ò˜ kÀÐÑOÔOÐOÐOÐOócó•—| t|¦«}tj|¦«}|j |¦«|_t t|¦« ||jd¬¦«|j |_ |S)NFr) Ú_get_checked_instancerÚtorchÚSizerÚexpandrrrÚ_validate_args)rrÚ _instanceÚnewrs €rr zOneHotCategorical.expand=s…ø€Ø×(Ò(Õ):¸IÑFÔFˆÝ”j Ñ-Ô-ˆ ØÔ,×3Ò3°KÑ@Ô@ˆÔÝ Õ Ñ%Ô%×.Ò.Ø ˜Ô)¸ð /ñ ô ð ð"Ô0ˆÔ؈ rcó&—|jj|i|¤ŽS©N)rÚ_new)rÚargsÚkwargss rr&zOneHotCategorical._newGs€Ø%ˆtÔ Ô% tÐ6¨vÐ6Ð6Ð6rcó—|jjSr%)rÚ_param©rs rr*zOneHotCategorical._paramJó€àÔ Ô'Ð'rcó—|jjSr%©rr r+s rr zOneHotCategorical.probsNó€àÔ Ô&Ð&rcó—|jjSr%)rr r+s rr zOneHotCategorical.logitsRr,rcó—|jjSr%r.r+s rÚmeanzOneHotCategorical.meanVr/rcóΗ|jj}| d¬¦«}tjj ||jd¬¦« |¦«S)Nr)Údim)Ú num_classes) rr ÚargmaxrÚnnÚ functionalÚone_hotÚshapeÚto)rr Úmodes rr<zOneHotCategorical.modeZsS€àÔ!Ô'ˆØ|Š| ˆ|Ñ#Ô#ˆÝŒxÔ"×*Ò*¨4¸U¼[ȼ_Ð*ÑMÔM×PÒPÐQVÑWÔWÐWrcó:—|jjd|jjz zS)Nér.r+s rÚvariancezOneHotCategorical.variance`s€àÔ Ô&¨!¨dÔ.?Ô.EÑ*EÑFÐFrcó—|jjSr%)rrr+s rrzOneHotCategorical.param_shapeds€àÔ Ô,Ð,rcóþ—tj|¦«}|jj}|jj}|j |¦«}tjj ||¦«  |¦«Sr%) rrrr Ú _num_eventsÚsampler7r8r9r;)rÚ sample_shaper Ú num_eventsÚindicess rrCzOneHotCategorical.samplehsf€Ý”z ,Ñ/Ô/ˆ ØÔ!Ô'ˆØÔ&Ô2ˆ ØÔ#×*Ò*¨<Ñ8Ô8ˆÝŒxÔ"×*Ò*¨7°JÑ?Ô?×BÒBÀ5ÑIÔIÐIrcó¤—|jr| |¦«| d¦«d}|j |¦«S)Nrr>)r!Ú_validate_sampleÚmaxrÚlog_prob)rÚvaluerFs rrJzOneHotCategorical.log_probosL€Ø Ô ð )Ø × !Ò ! %Ñ (Ô (Ð (Ø—)’)˜B‘-”- Ô"ˆØÔ ×)Ò)¨'Ñ2Ô2Ð2rcó4—|j ¦«Sr%)rÚentropyr+s rrMzOneHotCategorical.entropyus€ØÔ ×(Ò(Ñ*Ô*Ð*rcó"—|jd}tj||jj|jj¬¦«}| |fdt|j¦«zz|fz¦«}|r"|  |f|jz|fz¦«}|S)Nr)ÚdtypeÚdevice)r>) rrÚeyer*rOrPÚviewÚlenrr )rr ÚnÚvaluess rÚenumerate_supportz#OneHotCategorical.enumerate_supportxsŽ€Ø Ô ˜QÔ ˆÝ”˜1 D¤KÔ$5¸d¼kÔ>PÐQÑQÔQˆØ—’˜a˜T D­3¨tÔ/?Ñ+@Ô+@Ñ$@Ñ@ÀAÀ4ÑGÑHÔHˆØ ð CØ—]’] A 4¨$Ô*:Ñ#:¸a¸TÑ#AÑBÔBˆF؈ r)NNNr%)T)Ú__name__Ú __module__Ú __qualname__Ú__doc__rÚsimplexÚ real_vectorÚarg_constraintsr9ÚsupportÚhas_enumerate_supportrÚboolrr r&Úpropertyr*r r r2r<r?rrrrCrJrMrVÚ __classcell__)rs@rrrsXø€€€€€ððð<!,Ô 3¸{Ô?VÐWÐW€OØÔ!€GØ Ðð $Ø $Ø%)ð Pð Pà˜‰}ð P𘑠ð Pð˜d‘{ð Pð ð Pð Pð Pð Pð Pð Pððððððð7ð7ð7ðð(˜ð(ð(ð(ñ„Xð(ðð'vð'ð'ð'ñ„Xð'ðð(˜ð(ð(ð(ñ„Xð(ðð'fð'ð'ð'ñ„Xð'ððXfðXðXðXñ„XðXð ðG˜&ðGðGðGñ„XðGðð-˜UœZð-ð-ð-ñ„Xð-ð#- %¤*¡,¤,ðJðJðJðJð3ð3ð3ð +ð+ð+ððððððððrcóD—eZdZdZdZej¦«fdedefd„Z dS)r a  Creates a reparameterizable :class:`OneHotCategorical` distribution based on the straight- through gradient estimator from [1]. [1] Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computation (Bengio et al., 2013) TrDrcóx—| |¦«}|jj}||| ¦«z zSr%)rCrr Údetach)rrDÚsamplesr s rÚrsamplez(OneHotCategoricalStraightThrough.rsampleŒs6€Ø—+’+˜lÑ+Ô+ˆØÔ!Ô'ˆØ˜% %§,¢,¡.¤.Ñ0Ñ1Ð1rN) rWrXrYrZÚ has_rsamplerrrrrg©rrr r sW€€€€€ððð€Kà,6¨E¬J©L¬Lð2ð2 Eð2¸Vð2ð2ð2ð2ð2ð2r) rrÚtorch.distributionsrÚtorch.distributions.categoricalrÚ torch.distributions.distributionrÚ torch.typesrÚ__all__rr rirrúrosßðð € € € ØÐÐÐÐÐØ+Ð+Ð+Ð+Ð+Ð+Ø7Ð7Ð7Ð7Ð7Ð7Ø9Ð9Ð9Ð9Ð9Ð9ØÐÐÐÐÐð Ð BÐ C€ðpðpðpðpðp˜ ñpôpðpðf2ð2ð2ð2ð2Ð'8ñ2ô2ð2ð2ð2r