o iE@sddZddlZddlmZddlmZddlmZGdddZdd d d d d dddZ ddZ dS)zASR Interface module.N get_parser)dynamic_import)fill_missing_argsc@seZdZdZeddZededefddZdd Z dd d Z dd dZ ddZ ddZ eddZeddZddZddZddZd S) ASRInterfacez.ASR Interface for ESPnet model implementation.cCs|S)zAdd arguments to parser.parserrrM/home/ubuntu/.local/lib/python3.10/site-packages/espnet/nets/asr_interface.py add_arguments szASRInterface.add_argumentsidimodimcKs:dd}tjdi|}t||}t||j}||||S)aInitialize this class with python-level args. Args: idim (int): The number of an input feature dim. odim (int): The number of output vocab. Returns: ASRinterface: A new instance of ASRInterface. cSs t|ddS)NF)requiredrrrrr wraps z ASRInterface.build..wrapNr)argparse Namespacerr )clsr r kwargsrargsrrr builds     zASRInterface.buildcCtd)aeCompute loss for training. :param xs: For pytorch, batch of padded source sequences torch.Tensor (B, Tmax, idim) For chainer, list of source sequences chainer.Variable :param ilens: batch of lengths of source sequences (B) For pytorch, torch.Tensor For chainer, list of int :param ys: For pytorch, batch of padded source sequences torch.Tensor (B, Lmax) For chainer, list of source sequences chainer.Variable :return: loss value :rtype: torch.Tensor for pytorch, chainer.Variable for chainer z!forward method is not implementedNotImplementedErrorselfxsilensysrrr forward&szASRInterface.forwardNcCr)aaRecognize x for evaluation. :param ndarray x: input acouctic feature (B, T, D) or (T, D) :param namespace recog_args: argment namespace contraining options :param list char_list: list of characters :param torch.nn.Module rnnlm: language model module :return: N-best decoding results :rtype: list z#recognize method is not implementedrrx recog_args char_listrnnlmrrr recognize7 zASRInterface.recognizecCr)atBeam search implementation for batch. :param torch.Tensor x: encoder hidden state sequences (B, Tmax, Henc) :param namespace recog_args: argument namespace containing options :param list char_list: list of characters :param torch.nn.Module rnnlm: language model module :return: N-best decoding results :rtype: list z$Batch decoding is not supported yet.rrrrr recognize_batchCr%zASRInterface.recognize_batchcCr)a_Calculate attention. :param list xs: list of padded input sequences [(T1, idim), (T2, idim), ...] :param ndarray ilens: batch of lengths of input sequences (B) :param list ys: list of character id sequence tensor [(L1), (L2), (L3), ...] :return: attention weights (B, Lmax, Tmax) :rtype: float ndarray z2calculate_all_attentions method is not implementedrrrrr calculate_all_attentionsO z%ASRInterface.calculate_all_attentionscCr)ajCalculate CTC probability. :param list xs_pad: list of padded input sequences [(T1, idim), (T2, idim), ...] :param ndarray ilens: batch of lengths of input sequences (B) :param list ys: list of character id sequence tensor [(L1), (L2), (L3), ...] :return: CTC probabilities (B, Tmax, vocab) :rtype: float ndarray z1calculate_all_ctc_probs method is not implementedrrrrr calculate_all_ctc_probsZr(z$ASRInterface.calculate_all_ctc_probscCddlm}|S)zGet attention plot class.r)PlotAttentionReport)espnet.asr.asr_utilsr+)rr+rrr attention_plot_classe z!ASRInterface.attention_plot_classcCr*)zGet CTC plot class.r) PlotCTCReport)r,r/)rr/rrr ctc_plot_classlr.zASRInterface.ctc_plot_classcCr)zGet total subsampling factor.z6get_total_subsampling_factor method is not implementedrrrrr get_total_subsampling_factorssz)ASRInterface.get_total_subsampling_factorcCr)zEncode feature in `beam_search` (optional). Args: x (numpy.ndarray): input feature (T, D) Returns: torch.Tensor for pytorch, chainer.Variable for chainer: encoded feature (T, D) z encode method is not implementedr)rfeatrrr encodeyr%zASRInterface.encodecCr)zGet scorers for `beam_search` (optional). Returns: dict[str, ScorerInterface]: dict of `ScorerInterface` objects z"decoders method is not implementedrr1rrr scorersszASRInterface.scorers)NN)__name__ __module__ __qualname____doc__ staticmethodr classmethodintrrr$r&r'r)propertyr-r0r2r4r5rrrr r s$        rz'espnet.nets.pytorch_backend.e2e_asr:E2Ez2espnet.nets.pytorch_backend.e2e_asr_transducer:E2Ez3espnet.nets.pytorch_backend.e2e_asr_transformer:E2Ez1espnet.nets.pytorch_backend.e2e_asr_conformer:E2E)rnn transducer transformer conformerz'espnet.nets.chainer_backend.e2e_asr:E2Ez3espnet.nets.chainer_backend.e2e_asr_transformer:E2E)r>r@)pytorchchainercCs0t|t|t}t|tsJ|d|S)zImport ASR models dynamically. Args: module (str): module_name:class_name or alias in `predefined_asr` backend (str): NN backend. e.g., pytorch, chainer Returns: type: ASR class z does not implement ASRInterface)rpredefined_asrgetdict issubclassr)modulebackend model_classrrr dynamic_import_asrs rK) r9respnet.bin.asr_trainrespnet.utils.dynamic_importrespnet.utils.fill_missing_argsrrrDrKrrrr s"