add `SpeakerManager` tests

tests/inputs/test_speaker_encoder_config.json

@@ -0,0 +1,59 @@
+
+{
+    "run_name": "test_speaker_encoder",
+    "run_description": "test speaker encoder.",
+    "audio":{
+        // Audio processing parameters
+        "num_mels": 40,         // size of the mel spec frame.
+        "fft_size": 400,       // number of stft frequency levels. Size of the linear spectogram frame.
+        "sample_rate": 16000,   // DATASET-RELATED: wav sample-rate. If different than the original data, it is resampled.
+        "win_length": 400,     // stft window length in ms.
+        "hop_length": 160,      // stft window hop-lengh in ms.
+        "frame_length_ms": null,  // stft window length in ms.If null, 'win_length' is used.
+        "frame_shift_ms": null,   // stft window hop-lengh in ms. If null, 'hop_length' is used.
+        "preemphasis": 0.98,    // pre-emphasis to reduce spec noise and make it more structured. If 0.0, no -pre-emphasis.
+        "min_level_db": -100,   // normalization range
+        "ref_level_db": 20,     // reference level db, theoretically 20db is the sound of air.
+        "power": 1.5,           // value to sharpen wav signals after GL algorithm.
+        "griffin_lim_iters": 60,// #griffin-lim iterations. 30-60 is a good range. Larger the value, slower the generation.
+        // Normalization parameters
+        "signal_norm": true,    // normalize the spec values in range [0, 1]
+        "symmetric_norm": true, // move normalization to range [-1, 1]
+        "max_norm": 4.0,          // scale normalization to range [-max_norm, max_norm] or [0, max_norm]
+        "clip_norm": true,      // clip normalized values into the range.
+        "mel_fmin": 0.0,         // minimum freq level for mel-spec. ~50 for male and ~95 for female voices. Tune for dataset!!
+        "mel_fmax": 8000.0,        // maximum freq level for mel-spec. Tune for dataset!!
+        "do_trim_silence": true,  // enable trimming of slience of audio as you load it. LJspeech (false), TWEB (false), Nancy (true)
+        "trim_db": 60          // threshold for timming silence. Set this according to your dataset.
+    },
+    "reinit_layers": [],
+    "loss": "angleproto", // "ge2e" to use Generalized End-to-End loss and "angleproto" to use Angular Prototypical loss (new SOTA)
+    "grad_clip": 3.0, // upper limit for gradients for clipping.
+    "epochs": 1000, // total number of epochs to train.
+    "lr": 0.0001, // Initial learning rate. If Noam decay is active, maximum learning rate.
+    "lr_decay": false, // if true, Noam learning rate decaying is applied through training.
+    "warmup_steps": 4000, // Noam decay steps to increase the learning rate from 0 to "lr"
+    "tb_model_param_stats": false, // true, plots param stats per layer on tensorboard. Might be memory consuming, but good for debugging.
+    "steps_plot_stats": 10, // number of steps to plot embeddings.
+    "num_speakers_in_batch": 64, // Batch size for training. Lower values than 32 might cause hard to learn attention. It is overwritten by 'gradual_training'.
+    "num_utters_per_speaker": 10,  //
+    "num_loader_workers": 8,        // number of training data loader processes. Don't set it too big. 4-8 are good values.
+    "wd": 0.000001, // Weight decay weight.
+    "checkpoint": true, // If true, it saves checkpoints per "save_step"
+    "save_step": 1000, // Number of training steps expected to save traning stats and checkpoints.
+    "print_step": 20, // Number of steps to log traning on console.
+    "output_path": "", // DATASET-RELATED: output path for all training outputs.
+    "model": {
+        "input_dim": 40,
+        "proj_dim": 256,
+        "lstm_dim": 768,
+        "num_lstm_layers": 3,
+        "use_lstm_with_projection": true
+    },
+    "storage": {
+        "sample_from_storage_p": 0.66,  // the probability with which we'll sample from the DataSet in-memory storage
+        "storage_size": 15,   // the size of the in-memory storage with respect to a single batch
+        "additive_noise": 1e-5   // add very small gaussian noise to the data in order to increase robustness
+    },
+    "datasets":null
+}

tests/test_speakers_manager.py

@@ -0,0 +1,49 @@
+import os
+import unittest
+
+import numpy as np
+
+from tests import get_tests_input_path
+from TTS.tts.utils.speakers import SpeakerManager
+from TTS.utils.audio import AudioProcessor
+from TTS.utils.io import load_config
+
+encoder_config_path = os.path.join(get_tests_input_path(), "test_speaker_encoder_config.json")
+encoder_model_path = os.path.join(get_tests_input_path(), "dummy_speaker_encoder.pth.tar")
+sample_wav_path = os.path.join(get_tests_input_path(), "../data/ljspeech/wavs/LJ001-0001.wav")
+x_vectors_file_path = os.path.join(get_tests_input_path(), "../data/dummy_speakers.json")
+
+
+class SpeakerManagerTest(unittest.TestCase):
+    """Test SpeakerManager for loading embedding files and computing x_vectors from waveforms"""
+    @staticmethod
+    def test_speaker_embedding():
+        # load config
+        config = load_config(encoder_config_path)
+        config["audio"]["resample"] = True
+
+        # 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)
+        x_vector = manager.compute_x_vector(mel.T)
+        assert x_vector.shape[1] == 256
+
+    @staticmethod
+    def test_speakers_file_processing():
+        manager = SpeakerManager(x_vectors_file_path=x_vectors_file_path)
+        print(manager.num_speakers)
+        print(manager.x_vector_dim)
+        print(manager.clip_ids)
+        x_vector = manager.get_x_vector_by_clip(manager.clip_ids[0])
+        assert len(x_vector) == 256
+        x_vectors = manager.get_x_vectors_by_speaker(manager.speaker_ids[0])
+        assert len(x_vectors[0]) == 256
+        x_vector1 = manager.get_mean_x_vector(manager.speaker_ids[0], num_samples=2, randomize=True)
+        assert len(x_vector1) == 256
+        x_vector2 = manager.get_mean_x_vector(manager.speaker_ids[0], num_samples=2, randomize=False)
+        assert len(x_vector2) == 256
+        assert np.sum(np.array(x_vector1) - np.array(x_vector2)) != 0