第一次上传的一半

cluster/__init__.py

@@ -0,0 +1,29 @@
+import numpy as np
+import torch
+from sklearn.cluster import KMeans
+
+def get_cluster_model(ckpt_path):
+    checkpoint = torch.load(ckpt_path)
+    kmeans_dict = {}
+    for spk, ckpt in checkpoint.items():
+        km = KMeans(ckpt["n_features_in_"])
+        km.__dict__["n_features_in_"] = ckpt["n_features_in_"]
+        km.__dict__["_n_threads"] = ckpt["_n_threads"]
+        km.__dict__["cluster_centers_"] = ckpt["cluster_centers_"]
+        kmeans_dict[spk] = km
+    return kmeans_dict
+
+def get_cluster_result(model, x, speaker):
+    """
+        x: np.array [t, 256]
+        return cluster class result
+    """
+    return model[speaker].predict(x)
+
+def get_cluster_center_result(model, x,speaker):
+    """x: np.array [t, 256]"""
+    predict = model[speaker].predict(x)
+    return model[speaker].cluster_centers_[predict]
+
+def get_center(model, x,speaker):
+    return model[speaker].cluster_centers_[x]

cluster/train_cluster.py

@@ -0,0 +1,72 @@
+import os
+from glob import glob
+from pathlib import Path
+import torch
+import logging
+import argparse
+import torch
+import numpy as np
+from sklearn.cluster import KMeans, MiniBatchKMeans
+import tqdm
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger(__name__)
+import time
+import random
+
+def train_cluster(in_dir, n_clusters, use_minibatch=True, verbose=False):
+
+    logger.info(f"Loading features from {in_dir}")
+    features = []
+    nums = 0
+    for path in tqdm.tqdm(in_dir.glob("*.soft.pt")):
+        features.append(torch.load(path).squeeze(0).numpy().T)
+        # print(features[-1].shape)
+    features = np.concatenate(features, axis=0)
+    print(nums, features.nbytes/ 1024**2, "MB , shape:",features.shape, features.dtype)
+    features = features.astype(np.float32)
+    logger.info(f"Clustering features of shape: {features.shape}")
+    t = time.time()
+    if use_minibatch:
+        kmeans = MiniBatchKMeans(n_clusters=n_clusters,verbose=verbose, batch_size=4096, max_iter=80).fit(features)
+    else:
+        kmeans = KMeans(n_clusters=n_clusters,verbose=verbose).fit(features)
+    print(time.time()-t, "s")
+
+    x = {
+            "n_features_in_": kmeans.n_features_in_,
+            "_n_threads": kmeans._n_threads,
+            "cluster_centers_": kmeans.cluster_centers_,
+    }
+    print("end")
+
+    return x
+
+
+if __name__ == "__main__":
+
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--dataset', type=Path, default="./dataset/44k",
+                        help='path of training data directory')
+    parser.add_argument('--output', type=Path, default="./module/So-VITS/44k",
+                        help='path of model output directory')
+
+    args = parser.parse_args()
+
+    checkpoint_dir = args.output
+    dataset = args.dataset
+    n_clusters = 10000
+
+    ckpt = {}
+    for spk in os.listdir(dataset):
+        if os.path.isdir(dataset/spk):
+            print(f"train kmeans for {spk}...")
+            in_dir = dataset/spk
+            x = train_cluster(in_dir, n_clusters, verbose=False)
+            ckpt[spk] = x
+
+    checkpoint_path = checkpoint_dir / f"kmeans_{n_clusters}.pt"
+    checkpoint_path.parent.mkdir(exist_ok=True, parents=True)
+    torch.save(
+        ckpt,
+        checkpoint_path,
+    )

configs/config.json

@@ -0,0 +1,93 @@
+{
+  "train": {
+    "log_interval": 200,
+    "eval_interval": 800,
+    "seed": 1234,
+    "epochs": 10000,
+    "learning_rate": 0.0001,
+    "betas": [
+      0.8,
+      0.99
+    ],
+    "eps": 1e-09,
+    "batch_size": 6,
+    "fp16_run": false,
+    "lr_decay": 0.999875,
+    "segment_size": 10240,
+    "init_lr_ratio": 1,
+    "warmup_epochs": 0,
+    "c_mel": 45,
+    "c_kl": 1.0,
+    "use_sr": true,
+    "max_speclen": 512,
+    "port": "8001",
+    "keep_ckpts": 10
+  },
+  "data": {
+    "training_files": "filelists/train.txt",
+    "validation_files": "filelists/val.txt",
+    "max_wav_value": 32768.0,
+    "sampling_rate": 44100,
+    "filter_length": 2048,
+    "hop_length": 512,
+    "win_length": 2048,
+    "n_mel_channels": 80,
+    "mel_fmin": 0.0,
+    "mel_fmax": 22050
+  },
+  "model": {
+    "inter_channels": 192,
+    "hidden_channels": 192,
+    "filter_channels": 768,
+    "n_heads": 2,
+    "n_layers": 6,
+    "kernel_size": 3,
+    "p_dropout": 0.1,
+    "resblock": "1",
+    "resblock_kernel_sizes": [
+      3,
+      7,
+      11
+    ],
+    "resblock_dilation_sizes": [
+      [
+        1,
+        3,
+        5
+      ],
+      [
+        1,
+        3,
+        5
+      ],
+      [
+        1,
+        3,
+        5
+      ]
+    ],
+    "upsample_rates": [
+      8,
+      8,
+      2,
+      2,
+      2
+    ],
+    "upsample_initial_channel": 512,
+    "upsample_kernel_sizes": [
+      16,
+      16,
+      4,
+      4,
+      4
+    ],
+    "n_layers_q": 3,
+    "use_spectral_norm": false,
+    "gin_channels": 256,
+    "ssl_dim": 256,
+    "n_speakers": 200
+  },
+  "spk": {
+    "speaker0": 0
+  }
+}

dataset_raw/wav_structure.txt

@@ -0,0 +1,20 @@
+数据集准备
+
+raw
+├───speaker0
+│   ├───xxx1-xxx1.wav
+│   ├───...
+│   └───Lxx-0xx8.wav
+└───speaker1
+    ├───xx2-0xxx2.wav
+    ├───...
+    └───xxx7-xxx007.wav
+
+此外还需要编辑config.json
+
+"n_speakers": 10
+
+"spk":{
+    "speaker0": 0,
+    "speaker1": 1,
+}

face_detection/__init__.py

@@ -0,0 +1,7 @@
+# -*- coding: utf-8 -*-
+
+__author__ = """Adrian Bulat"""
+__email__ = '[email protected]'
+__version__ = '1.0.1'
+
+from .api import FaceAlignment, LandmarksType, NetworkSize

face_detection/api.py

@@ -0,0 +1,79 @@
+from __future__ import print_function
+import os
+import torch
+from torch.utils.model_zoo import load_url
+from enum import Enum
+import numpy as np
+import cv2
+try:
+    import urllib.request as request_file
+except BaseException:
+    import urllib as request_file
+
+from .models import FAN, ResNetDepth
+from .utils import *
+
+
+class LandmarksType(Enum):
+    """Enum class defining the type of landmarks to detect.
+
+    ``_2D`` - the detected points ``(x,y)`` are detected in a 2D space and follow the visible contour of the face
+    ``_2halfD`` - this points represent the projection of the 3D points into 3D
+    ``_3D`` - detect the points ``(x,y,z)``` in a 3D space
+
+    """
+    _2D = 1
+    _2halfD = 2
+    _3D = 3
+
+
+class NetworkSize(Enum):
+    # TINY = 1
+    # SMALL = 2
+    # MEDIUM = 3
+    LARGE = 4
+
+    def __new__(cls, value):
+        member = object.__new__(cls)
+        member._value_ = value
+        return member
+
+    def __int__(self):
+        return self.value
+
+ROOT = os.path.dirname(os.path.abspath(__file__))
+
+class FaceAlignment:
+    def __init__(self, landmarks_type, network_size=NetworkSize.LARGE,
+                 device='cuda', flip_input=False, face_detector='sfd', verbose=False):
+        self.device = device
+        self.flip_input = flip_input
+        self.landmarks_type = landmarks_type
+        self.verbose = verbose
+
+        network_size = int(network_size)
+
+        if 'cuda' in device:
+            torch.backends.cudnn.benchmark = True
+
+        # Get the face detector
+        face_detector_module = __import__('face_detection.detection.' + face_detector,
+                                          globals(), locals(), [face_detector], 0)
+        self.face_detector = face_detector_module.FaceDetector(device=device, verbose=verbose)
+
+    def get_detections_for_batch(self, images):
+        images = images[..., ::-1]
+        detected_faces = self.face_detector.detect_from_batch(images.copy())
+        results = []
+
+        for i, d in enumerate(detected_faces):
+            if len(d) == 0:
+                results.append(None)
+                continue
+            d = d[0]
+            d = np.clip(d, 0, None)
+
+            x1, y1, x2, y2 = map(int, d[:-1])
+            results.append((x1, y1, x2, y2))
+
+        return results

face_detection/detection/__init__.py

@@ -0,0 +1 @@
+from .core import FaceDetector