coqui-ai
/
TTS
mirror of https://github.com/coqui-ai/TTS.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #1054 from WeberJulian/partial_embedding_compute 

Partial embedding compute
pull/1227/head
Eren Gölge 2022-02-06 20:13:55 +01:00 committed by GitHub
parent 2ca2c8c431 c7f5e005e1
commit 44c7d1a826
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 9 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
TTS/bin/compute_embeddings.py
View File

@ -29,6 +29,7 @@ parser.add_argument(
help="Path to dataset config file.", help="Path to dataset config file.",
) )
parser.add_argument("output_path", type=str, help="path for output speakers.json and/or speakers.npy.") parser.add_argument("output_path", type=str, help="path for output speakers.json and/or speakers.npy.")
parser.add_argument("--old_file", type=str, help="Previous speakers.json file, only compute for new audios.", default=None)
parser.add_argument("--use_cuda", type=bool, help="flag to set cuda.", default=True) parser.add_argument("--use_cuda", type=bool, help="flag to set cuda.", default=True)
parser.add_argument("--eval", type=bool, help="compute eval.", default=True) parser.add_argument("--eval", type=bool, help="compute eval.", default=True)
@ -40,7 +41,7 @@ meta_data_train, meta_data_eval = load_tts_samples(c_dataset.datasets, eval_spli
wav_files = meta_data_train + meta_data_eval wav_files = meta_data_train + meta_data_eval
speaker_manager = SpeakerManager( speaker_manager = SpeakerManager(
encoder_model_path=args.model_path, encoder_config_path=args.config_path, use_cuda=args.use_cuda encoder_model_path=args.model_path, encoder_config_path=args.config_path, d_vectors_file_path=args.old_file, use_cuda=args.use_cuda
) )
# compute speaker embeddings # compute speaker embeddings
@ -52,11 +53,15 @@ for idx, wav_file in enumerate(tqdm(wav_files)):
else: else:
speaker_name = None speaker_name = None
wav_file_name = os.path.basename(wav_file)
if args.old_file is not None and wav_file_name in speaker_manager.clip_ids:
# get the embedding from the old file
embedd = speaker_manager.get_d_vector_by_clip(wav_file_name)
else:
# extract the embedding # extract the embedding
embedd = speaker_manager.compute_d_vector_from_clip(wav_file) embedd = speaker_manager.compute_d_vector_from_clip(wav_file)
# create speaker_mapping if target dataset is defined # create speaker_mapping if target dataset is defined
wav_file_name = os.path.basename(wav_file)
speaker_mapping[wav_file_name] = {} speaker_mapping[wav_file_name] = {}
speaker_mapping[wav_file_name]["name"] = speaker_name speaker_mapping[wav_file_name]["name"] = speaker_name
speaker_mapping[wav_file_name]["embedding"] = embedd speaker_mapping[wav_file_name]["embedding"] = embedd