@i SrSSKrSSKJrJr SSKrSSKrSSKJ r Sr SSK Jr \R"\5r"SS 5r"S S \ (a\ R&5rg\5rg!\ a Sr \R"S5 Nbf=f) z5Pattern encoder with learnable preference projection.N)ListUnionTFz9PyTorch not available, PreferenceEncoder will be disabled)SentenceTransformerc`\rSrSrSrS S\4SjjrS\\\\4S\ R4Sjr Sr g ) PatternEncoderz:Base pattern encoder using sentence transformers (frozen). model_namec[U5UlURR5Ul[R SUR35 g)zZInitialize pattern encoder. Args: model_name: Name of the sentence transformer model z*PatternEncoder initialized with dimension N)rmodel get_sentence_embedding_dimension embedding_dimloggerinfo)selfr s D/home/ubuntu/codebase/yexijia/保研/iCoLoc/src/embedding/encoder.py__init__PatternEncoder.__init__s@ )4 !ZZHHJ @ASAS@TUVtextreturnc[U[5(aURRU5$URRU5$)zEncode pattern text to vector. Args: text: Pattern text (e.g., "Park Zoo") or list of patterns Returns: Vector representation(s) ) isinstancestrr encode)rrs rrPatternEncoder.encode!s; dC ::$$T* *::$$T* *r)r r N)z&sentence-transformers/all-MiniLM-L6-v2) __name__ __module__ __qualname____firstlineno____doc__rrrrnpndarrayr__static_attributes__rrrrs7DW3W +5d3i0 +RZZ +rrc^\rSrSrSrS S\S\S\4U4SjjjrSrS\RS \R4S jr S\ \ R\R4S \ R4S jrS rU=r$)PreferenceEncoder0zLearnable preference encoder for projecting embeddings to preference space. Architecture: Linear(dim -> hidden) -> ReLU -> Linear(hidden -> dim) -> Normalize input_dim hidden_dim output_dimc>[(d [S5e[[U]5 UcUnXlX lX0l[R"X5Ul [R"5Ul [R"X#5Ul UR5 [R!SUSUSU35 g)zInitialize preference encoder. Args: input_dim: Input embedding dimension hidden_dim: Hidden layer dimension output_dim: Output dimension (default: same as input_dim) z)PyTorch is required for PreferenceEncoderNzPreferenceEncoder initialized: z -> )TORCH_AVAILABLE ImportErrorsuperr&rr(r)r*nnLinearfc1ReLUrelufc2_initialize_weightsrr)rr(r)r* __class__s rrPreferenceEncoder.__init__7sIJ J /1  "J"$$99Y3GGI 99Z4   " 5i[ZLPTU_T`abrc[RRURR5 [RR URR 5 [RRURR5 [RR URR 5 g)zInitialize network weights.N)r/initxavier_uniform_r1weightzeros_biasr4)rs rr5%PreferenceEncoder._initialize_weightsUsf 0 txx}}% 0 txx}}%rxrcURU5nURU5nURU5n[RR R USSS9nU$)zForward pass. Args: x: Input tensor of shape (batch_size, input_dim) Returns: Projected and normalized tensor of shape (batch_size, output_dim) )pdim)r1r3r4torchr/ functional normalize)rr?s rforwardPreferenceEncoder.forward\sS HHQK IIaL HHQK HH   ) )!qa ) 8rc[U[R5(a[R"U5n[ UR 5S:XaURS5nUR5 [R"5 URU5nSSS5 WR SS:XaURS5R5$UR5$!,(df  NP=f)zEncode input to preference space (for inference). Args: x: Input array or tensor Returns: Projected vector as numpy array rBrN) rr!r"rE FloatTensorlenshape unsqueezeevalno_gradrHsqueezenumpy)rr?outputs rrPreferenceEncoder.encodens a $ $!!!$A qww<1  AA ]]_\\!_F <<?a >>!$**, ,<<> ! _s C!! C/)r1r4r)r(r*r3)iN)rrrrr intrr5rETensorrHrr!r"rr# __classcell__)r6s@rr&r&0s{ c#ccPScc<&%,,$"bjj%,,67"BJJ""rr&)r rRr!typingrrloggingrEtorch.nnr/r,r-warningsentence_transformersr getLoggerrrrModuleobjectr&r$rrras;QO 6   8 $++8T"_ T"&T"KQO OOOPQs A&&BB