dj" dZddlZddlZddlmZddlmZddlmZ ddl m Z ddl m Zddl mZdd l mZdd l mZdd l mZej(ZGd d ej,ZGddej,Zej2dej4eej4ej6gej2dej6ej8ej4e j:gej2dej6gej<dej4e j:gdfZej@GddZ!GddZ"y)zMediaPipe text embedder task.N)Optional)embedding_result)embedding_result_ccosine_similarity base_options)base_options_c)mediapipe_c_bindings)mediapipe_c_utils)serial_dispatchercHeZdZdZdej fdej fgZy)_EmbedderOptionsCzC struct for embedder options. l2_normalizequantizeN)__name__ __module__ __qualname____doc__ctypesc_bool_fields_l/root/repos/pocs/video-evals/.venv/lib/python3.12/site-packages/mediapipe/tasks/python/text/text_embedder.pyrrs&&v}}%6==!(rrc4eZdZdZdej fdefgZy)_TextEmbedderOptionsCz#C struct for text embedder options.r embedder_optionsN)rrrrbase_options_c_module BaseOptionsCrrrrrrr(s&+,99:,-(rrMpTextEmbedderCreateMpTextEmbedderEmbedMpTextEmbedderCloseMpTextEmbedderCloseResultcdeZdZUdZej ed<dZee ed<dZ ee ed<de fdZ y)TextEmbedderOptionsaOptions for the text embedder task. Attributes: base_options: Base options for the text embedder task. l2_normalize: Whether to normalize the returned feature vector with L2 norm. Use this option only if the model does not already contain a native L2_NORMALIZATION TF Lite Op. In most cases, this is already the case and L2 norm is thus achieved through TF Lite inference. quantize: Whether the returned embedding should be quantized to bytes via scalar quantization. Embeddings are implicitly assumed to be unit-norm and therefore any dimension is guaranteed to have a value in [-1.0, 1.0]. Use the l2_normalize option if this is not the case. r Nrrreturnc|jj}t|j |jnd|j |jnd}t ||S)z/Generates a ctypes TextEmbedderOptionsC object.F)rr)r r)r to_ctypesrrrr)selfr embedder_options_cs rr)zTextEmbedderOptions.to_ctypesbsb&&002N*    (&& "&--";  !#6H r) rrrrbase_options_module BaseOptions__annotations__rrboolrrr)rrrr&r&Os> $///!%,%!(HTN! . rr&c<eZdZUdZej ed<ejed<dej dejfdZ e de ddfd Z e d e ddfd Zd e dej fd Ze dej$dej$defdZdZdZdZdZy) TextEmbedderaoClass that performs embedding extraction on text. This API expects a TFLite model with TFLite Model Metadata that contains the mandatory (described below) input tensors and output tensors. Metadata should contain the input process unit for the model's Tokenizer as well as input / output tensor metadata. Input tensors: (kTfLiteInt32) - 3 input tensors of size `[batch_size x bert_max_seq_len]` with names "ids", "mask", and "segment_ids" representing the input ids, mask ids, and segment ids respectively. - or 1 input tensor of size `[batch_size x max_seq_len]` representing the input ids. At least one output tensor with: (kTfLiteFloat32) - `N` components corresponding to the `N` dimensions of the returned feature vector for this output layer. - Either 2 or 4 dimensions, i.e. `[1 x N]` or `[1 x 1 x 1 x N]`. _lib_handlelibhandlec ||_||_yN)r2_embedder_handle)r*r4r5s r__init__zTextEmbedder.__init__s DI"Dr model_pathr'chtj|}t|}|j|S)aCreates an `TextEmbedder` object from a TensorFlow Lite model and the default `TextEmbedderOptions`. Args: model_path: Path to the model. Returns: `TextEmbedder` object that's created from the model file and the default `TextEmbedderOptions`. Raises: ValueError: If failed to create `TextEmbedder` object from the provided file such as invalid file path. RuntimeError: If other types of error occurred. )model_asset_pathr)r,r-r&create_from_options)clsr:r optionss rcreate_from_model_pathz#TextEmbedder.create_from_model_paths/ '22JOL!|r?r4ctypes_optionsembedder_handles rr=z TextEmbedder.create_from_optionssc  2 23E FC&&(Noo'O ^$fll?&C C 88rtextcZtj}|jj|j|j dt j|tjj|}|jjt j||S)a4Performs text embedding extraction on the provided text. Args: text: The input text. Returns: An embedding result object that contains a list of embeddings. Raises: ValueError: If any of the input arguments is invalid. RuntimeError: If text embedder failed to run. zutf-8) embedding_result_c_moduleEmbeddingResultCr2r"r8encoderrEembedding_result_moduleEmbeddingResult from_ctypesr$)r*rH ctypes_result python_results rembedzTextEmbedder.embeds .>>@MII!!  G ]# //;;    II'' ](CD ruvc.tj||S)a?Utility function to compute cosine similarity between two embedding entries. May return an InvalidArgumentError if e.g. the feature vectors are of different types (quantized vs. float), have different sizes, or have a an L2-norm of 0. Args: u: An embedding entry. v: An embedding entry. Returns: The cosine similarity for the two embeddings. Raises: ValueError: May return an error if e.g. the feature vectors are of different types (quantized vs. float), have different sizes, or have an L2-norm of 0. r)r>rSrTs rrzTextEmbedder.cosine_similaritys0  . .q! 44rc|jrG|jj|jd|_|jjyy)z'Shuts down the MediaPipe task instance.N)r8r2r#closer*s rrWzTextEmbedder.closes>  ii##D$9$9:"d iioorc|S)z1Returns `self` upon entering the runtime context.rrXs r __enter__zTextEmbedder.__enter__s Krc$|jy)zFShuts down the MediaPipe task instance on exit of the context manager.NrW)r*unused_exc_typeunused_exc_valueunused_tracebacks r__exit__zTextEmbedder.__exit__s JJLrc$|jyr7r\rXs r__del__zTextEmbedder.__del__s JJLrN)rrrrr SerialDispatcherr.rrDr9 classmethodstrr@r&r=rMrNrR EmbeddingfloatrrWrZr`rbrrrr1r1ps* *** ??#  - -#oo#,c,n,,&9(;999, ..<5 * *5 ! * *5 552rr1)#rr dataclassestypingr,mediapipe.tasks.python.components.containersrrMrrJ'mediapipe.tasks.python.components.utilsrmediapipe.tasks.python.corer r,r rr r r rNTextEmbedderResult StructurerrCStatusFunctionPOINTERrDc_char_prK CFunctionrC dataclassr&r1rrrrts\$ dhEKO<99,<<((F,,&%% FNN0 1 FNN6?? + &%% OO OO FNN4EE F &%% OO   # 1BB CD -< @TTr