From 631aec6e880ca6f941f2a4ee92f113482d0cd8fa Mon Sep 17 00:00:00 2001 From: Edresson Casanova Date: Mon, 7 Mar 2022 19:25:04 -0300 Subject: [PATCH] Active evaluation in speaker encoder test and use multispeaker dataset for this test --- tests/aux_tests/test_speaker_encoder_train.py | 15 +++++++++------ 1 file changed, 9 insertions(+), 6 deletions(-) diff --git a/tests/aux_tests/test_speaker_encoder_train.py b/tests/aux_tests/test_speaker_encoder_train.py index 40867e53..ddf526c7 100644 --- a/tests/aux_tests/test_speaker_encoder_train.py +++ b/tests/aux_tests/test_speaker_encoder_train.py @@ -4,14 +4,14 @@ import shutil from tests import get_device_id, get_tests_output_path, run_cli from TTS.config.shared_configs import BaseAudioConfig -from TTS.encoder.speaker_encoder_config import SpeakerEncoderConfig +from TTS.encoder.configs.speaker_encoder_config import SpeakerEncoderConfig def run_test_train(): command = ( f"CUDA_VISIBLE_DEVICES='{get_device_id()}' python TTS/bin/train_encoder.py --config_path {config_path} " f"--coqpit.output_path {output_path} " - "--coqpit.datasets.0.name ljspeech " + "--coqpit.datasets.0.name ljspeech_test " "--coqpit.datasets.0.meta_file_train metadata.csv " "--coqpit.datasets.0.meta_file_val metadata.csv " "--coqpit.datasets.0.path tests/data/ljspeech " @@ -24,13 +24,16 @@ output_path = os.path.join(get_tests_output_path(), "train_outputs") config = SpeakerEncoderConfig( batch_size=4, - num_classes_in_batch=1, - num_utter_per_class=10, - num_loader_workers=0, - max_train_step=2, + num_classes_in_batch=2, + num_utter_per_class=2, + eval_num_classes_in_batch=2, + eval_num_utter_per_class=2, + num_loader_workers=1, + epochs=2, print_step=1, save_step=1, print_eval=True, + run_eval=True, audio=BaseAudioConfig(num_mels=80), ) config.audio.do_trim_silence = True