update

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,
+}

filelists/test.txt

@@ -0,0 +1,4 @@
+./dataset/44k/taffy/000562.wav
+./dataset/44k/nyaru/000011.wav
+./dataset/44k/nyaru/000008.wav
+./dataset/44k/taffy/000563.wav

filelists/train.txt

@@ -0,0 +1,15 @@
+./dataset/44k/taffy/000549.wav
+./dataset/44k/nyaru/000004.wav
+./dataset/44k/nyaru/000006.wav
+./dataset/44k/taffy/000551.wav
+./dataset/44k/nyaru/000009.wav
+./dataset/44k/taffy/000561.wav
+./dataset/44k/nyaru/000001.wav
+./dataset/44k/taffy/000553.wav
+./dataset/44k/nyaru/000002.wav
+./dataset/44k/taffy/000560.wav
+./dataset/44k/taffy/000557.wav
+./dataset/44k/nyaru/000005.wav
+./dataset/44k/taffy/000554.wav
+./dataset/44k/taffy/000550.wav
+./dataset/44k/taffy/000559.wav

filelists/val.txt

@@ -0,0 +1,4 @@
+./dataset/44k/nyaru/000003.wav
+./dataset/44k/nyaru/000007.wav
+./dataset/44k/taffy/000558.wav
+./dataset/44k/taffy/000556.wav

pretrain/checkpoint_best_legacy_500.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f54b40fd2802423a5643779c4861af1e9ee9c1564dc9d32f54f20b5ffba7db96
+size 189507909

pretrain/meta.py

@@ -0,0 +1,39 @@
+def download_dict():
+    return {
+        "vec768l12": {
+            "url": "https://ibm.ent.box.com/shared/static/z1wgl1stco8ffooyatzdwsqn2psd9lrr",
+            "output": "./pretrain/checkpoint_best_legacy_500.pt"
+        },
+        "vec256l9": {
+            "url": "https://ibm.ent.box.com/shared/static/z1wgl1stco8ffooyatzdwsqn2psd9lrr",
+            "output": "./pretrain/checkpoint_best_legacy_500.pt"
+        },
+        "hubertsoft": {
+            "url": "https://github.com/bshall/hubert/releases/download/v0.1/hubert-soft-0d54a1f4.pt",
+            "output": "./pretrain/hubert-soft-0d54a1f4.pt"
+        },
+        "whisper-ppg-small": {
+            "url": "https://openaipublic.azureedge.net/main/whisper/models/9ecf779972d90ba49c06d968637d720dd632c55bbf19d441fb42bf17a411e794/small.pt",
+            "output": "./pretrain/small.pt"
+        },
+        "whisper-ppg": {
+            "url": "https://openaipublic.azureedge.net/main/whisper/models/345ae4da62f9b3d59415adc60127b97c714f32e89e936602e85993674d08dcb1/medium.pt",
+            "output": "./pretrain/medium.pt"
+        },
+        "whisper-ppg-large": {
+            "url": "https://openaipublic.azureedge.net/main/whisper/models/81f7c96c852ee8fc832187b0132e569d6c3065a3252ed18e56effd0b6a73e524/large-v2.pt",
+            "output": "./pretrain/large-v2.pt"
+        }
+    }
+
+
+def get_speech_encoder(config_path="configs/config.json"):
+    import json
+
+    with open(config_path, "r") as f:
+        data = f.read()
+        config = json.loads(data)
+        speech_encoder = config["model"]["speech_encoder"]
+        dict = download_dict()
+
+        return dict[speech_encoder]["url"], dict[speech_encoder]["output"]

trained/nahida/nahida.json

@@ -0,0 +1,96 @@
+{
+  "train": {
+    "log_interval": 200,
+    "eval_interval": 800,
+    "seed": 1234,
+    "epochs": 10000,
+    "learning_rate": 0.00015,
+    "betas": [
+      0.8,
+      0.99
+    ],
+    "eps": 1e-09,
+    "batch_size": 12,
+    "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": 3,
+    "all_in_mem": false
+  },
+  "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": 1,
+    "speech_encoder": "vec256l9",
+    "speaker_embedding": false
+  },
+  "spk": {
+    "nahida": 0
+  }
+}

trained/nahida/nahida_G_40000.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fec9506f8a762516e5321bcaedc72b6891d61a631e82ac596fe4e16e45b4652a
+size 542178077