o j% @sddlZddlZddlmZmZmZmZddlZddlZ ddl Z ddl m Z ddl mZddlmZGdddeZdd Zd d Zd d Zdde dedededef ddZdefddZdS)N)AnyDictListUnion)Coqpit)*get_from_config_or_model_args_with_default)EmbeddingManagerc seZdZdZ      ddeeedededed ed ef fd d Ze d dZ e ddZ defddZ e ddddeeeeefddfddZZS)SpeakerManagerujManage the speakers for multi-speaker 🐸TTS models. Load a datafile and parse the information in a way that can be queried by speaker or clip. There are 3 different scenarios considered: 1. Models using speaker embedding layers. The datafile only maps speaker names to ids used by the embedding layer. 2. Models using d-vectors. The datafile includes a dictionary in the following format. :: { 'clip_name.wav':{ 'name': 'speakerA', 'embedding'[] }, ... } 3. Computing the d-vectors by the speaker encoder. It loads the speaker encoder model and computes the d-vectors for a given clip or speaker. Args: d_vectors_file_path (str, optional): Path to the metafile including x vectors. Defaults to "". speaker_id_file_path (str, optional): Path to the metafile that maps speaker names to ids used by TTS models. Defaults to "". encoder_model_path (str, optional): Path to the speaker encoder model file. Defaults to "". encoder_config_path (str, optional): Path to the spealer encoder config file. Defaults to "". Examples: >>> # load audio processor and speaker encoder >>> ap = AudioProcessor(**config.audio) >>> manager = SpeakerManager(encoder_model_path=encoder_model_path, encoder_config_path=encoder_config_path) >>> # load a sample audio and compute embedding >>> waveform = ap.load_wav(sample_wav_path) >>> mel = ap.melspectrogram(waveform) >>> d_vector = manager.compute_embeddings(mel.T) NF data_itemsd_vectors_file_pathspeaker_id_file_pathencoder_model_pathencoder_config_pathuse_cudacs0tj|||||d|r|j|dddSdS)N)embedding_file_path id_file_pathrrr speaker_name parse_key)super__init__set_ids_from_data)selfr r r rrr __class__I/home/kuhnn/.local/lib/python3.10/site-packages/TTS/tts/utils/speakers.pyr6s zSpeakerManager.__init__cCs t|jSN)len name_to_idrrrr num_speakersJs zSpeakerManager.num_speakerscCst|jSr)listr keysr!rrr speaker_namesNszSpeakerManager.speaker_namesreturncCs|jSrr r!rrr get_speakersRszSpeakerManager.get_speakersconfigrsamplescCsd}t|ddr-|rt|d}t|ddrtt|ddd}t|ddr-tt|ddd}t|ddrEt}t|d drEtt|d dd }|S) aHInitialize a speaker manager from config Args: config (Coqpit): Config object. samples (Union[List[List], List[Dict]], optional): List of data samples to parse out the speaker names. Defaults to None. Returns: SpeakerEncoder: Speaker encoder object. Nuse_speaker_embeddingF)r speaker_file)r speakers_fileuse_d_vector_file d_vector_file)r )rr )r)r*speaker_managerrrrinit_from_configUs&         zSpeakerManager.init_from_config)Nr r r r Fr)__name__ __module__ __qualname____doc__rrstrboolrpropertyr"r%r( staticmethodrrr1 __classcell__rrrrr s8)   2r cCsZtjtj|d}tj|d}t|j}||r|S||r&|Std|)zFind the speakers.json under the given path or the above it. Intended to band aid the different paths returned in restored and continued training. speakers.jsonz" [!] `speakers.json` not found in ) ospathjoindirnamefsspec get_mapperfsexistsFileNotFoundError)r= path_restore path_continuerBrrr_set_file_pathws   rGcCs\tj|ddkr |}nt|}t|d }t|WdS1s'wYdS)z)Loads speaker mapping if already present.z.jsonrN)r<r=splitextrGr@openjsonload)out_path json_filefrrrload_speaker_mappings $rQcCsV|dur)t|}t|d}tj||ddWddS1s"wYdSdS)z)Saves speaker mapping if not yet present.Nw)indent)rGr@rKrLdump)rNspeaker_mappingspeakers_json_pathrPrrrsave_speaker_mappings "rXcdata restore_pathrNr&csdt|jr|durj|dd|rVt|}|jr;tj|s5tdtj|j s/t d |j  |n@|jsUj } |tfdd|DsUJdn%|jrc|j rc |j n|jrk|j skd |jr{d |vr{|jr{ |jjd krtd jd j |rtj|d}td|d|jr|j r|S|S)a Initiate a `SpeakerManager` instance by the provided config. Args: c (Coqpit): Model configuration. restore_path (str): Path to a previous training folder. data (List): Data samples used in training to infer speakers from. It must be provided if speaker embedding layers is used. Defaults to None. out_path (str, optional): Save the generated speaker IDs to a output path. Defaults to None. Returns: SpeakerManager: initialized and ready to use instance. NrrzXWARNING: speakers.json was not found in restore_path, trying to use CONFIG.d_vector_filezaYou must copy the file speakers.json to restore_path, or set a valid file in CONFIG.d_vector_filec3s|]}|jvVqdSrr').0speakerr0rr s  z&get_speaker_manager..z> [!] You cannot introduce new speakers to a pre-trained model.zNuse_d_vector_file is True, so you need pass a external speaker embedding file.r-rz1 > Speaker manager is loaded with {} speakers: {}z, r;z > Saving `speakers.json` to .)r r+rrGr.r<r=rCprintr/ RuntimeErrorload_embeddings_from_filer load_ids_from_fileallr-r"formatr>save_embeddings_to_filesave_ids_to_file)rYrZr[rNr-speaker_ids_from_data out_file_pathrr^rget_speaker_managersX              rkitemscstdd|DtfddD}tfddD}d|tfdd|D}|tj|}t|S)NcSsg|]}|dqS)rr)r\itemrrr z0get_speaker_balancer_weights..csg|]}|qSr)indexr\l)unique_speaker_namesrrrnscs"g|] }tt|kdqS)r)rnpwhererq)r%rrrns"g?csg|]}|qSrrrq)weight_speakerrrrnro) rtarrayuniquetolistlinalgnormtorch from_numpyfloat)rl speaker_ids speaker_countdataset_samples_weightr)r%rsrvrget_speaker_balancer_weightssr)NNN)rLr<typingrrrrr@numpyrtr|coqpitr TTS.configrTTS.tts.utils.managersrr rGrQrXr6rkr#rrrrrs   i  ?