Upload 29 files

.gitattributes

@@ -1 +1,23 @@
 * text=auto eol=lf
+pre_trained_model/768l12/D_0.pth filter=lfs diff=lfs merge=lfs -text
+pre_trained_model/768l12/G_0.pth filter=lfs diff=lfs merge=lfs -text
+pre_trained_model/768l12/vol_emb/D_0.pth filter=lfs diff=lfs merge=lfs -text
+pre_trained_model/768l12/vol_emb/G_0.pth filter=lfs diff=lfs merge=lfs -text
+pre_trained_model/D_0.pth filter=lfs diff=lfs merge=lfs -text
+pre_trained_model/diffusion/768l12/max100/model_0.pt filter=lfs diff=lfs merge=lfs -text
+pre_trained_model/diffusion/768l12/model_0.pt filter=lfs diff=lfs merge=lfs -text
+pre_trained_model/diffusion/hubertsoft/model_0.pt filter=lfs diff=lfs merge=lfs -text
+pre_trained_model/diffusion/whisper-ppg/model_0.pt filter=lfs diff=lfs merge=lfs -text
+pre_trained_model/G_0.pth filter=lfs diff=lfs merge=lfs -text
+pre_trained_model/hubertsoft/D_0.pth filter=lfs diff=lfs merge=lfs -text
+pre_trained_model/hubertsoft/G_0.pth filter=lfs diff=lfs merge=lfs -text
+pre_trained_model/tiny/vec768l12_vol_emb/D_0.pth filter=lfs diff=lfs merge=lfs -text
+pre_trained_model/tiny/vec768l12_vol_emb/G_0.pth filter=lfs diff=lfs merge=lfs -text
+pre_trained_model/whisper-ppg/D_0.pth filter=lfs diff=lfs merge=lfs -text
+pre_trained_model/whisper-ppg/G_0.pth filter=lfs diff=lfs merge=lfs -text
+pretrain/checkpoint_best_legacy_500.pt filter=lfs diff=lfs merge=lfs -text
+pretrain/fcpe.pt filter=lfs diff=lfs merge=lfs -text
+pretrain/hubert-soft-0d54a1f4.pt filter=lfs diff=lfs merge=lfs -text
+pretrain/medium.pt filter=lfs diff=lfs merge=lfs -text
+pretrain/nsf_hifigan/model filter=lfs diff=lfs merge=lfs -text
+pretrain/rmvpe.pt filter=lfs diff=lfs merge=lfs -text

pre_trained_model/768l12/D_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:60b6936d55d2cfaa717033eafe9d98dbe44d322e6adaf7be7c1c5a835ebb7177
+size 187027770

pre_trained_model/768l12/G_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9d3e408786013590bb3574ade2831ab62c989d303834742fe73ca8d5552d2f03
+size 209268661

pre_trained_model/768l12/vol_emb/D_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2506a5457b2f6133563418ab0429463d3ca9788cc2575979d3839a8699b3c158
+size 187027770

pre_trained_model/768l12/vol_emb/G_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:176d6e0dde4ba7de5ebfe1d01a1afebb161ec9945ea5cbee3131f187f53b39be
+size 209270847

pre_trained_model/D_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:22007b495c822f9b3c90eb6c225844f4d203bea7f5c214bcd9b5dbc2a247ed7e
+size 187018271

pre_trained_model/G_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ea40d6066461974c36f8984f956a7d2cafcce517a3bc6222a8877fab419c0179
+size 180619621

pre_trained_model/diffusion/768l12/max100/model_0.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a4500926b38f792692d3985e9489127592879344d9b3850c8a06d506c18b17f
+size 325929182

pre_trained_model/diffusion/768l12/model_0.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:409452a27ab310f7a5897844d003d372a7357cc91c4a43562584a1714518cdf9
+size 220895384

pre_trained_model/diffusion/hubertsoft/model_0.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0e4b0e92dbc764456d39985f1e168117edd6eb3cea4a6afc7166fcd701c83ed8
+size 220371096

pre_trained_model/diffusion/whisper-ppg/model_0.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0cdbc3e2dbc677b9c6fa5c96865cabd558e8f50afd76abd37460333e7153dbe6
+size 221156297

pre_trained_model/hubertsoft/D_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5424bd5508fb198025d359f727357ec2421ca65ef3b86b692242ece0e4c370d6
+size 187027770

pre_trained_model/hubertsoft/G_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a78c3c6f58a11f16c0ba92a8e51c9532697ee598771b729f4c3779afe130ab43
+size 152358837

pre_trained_model/tiny/vec768l12_vol_emb/D_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3f653339e71ddc481375b5c6118d04ace2b2878490e6b0729cccb91beb0d2ffc
+size 187029190

pre_trained_model/tiny/vec768l12_vol_emb/G_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b4a3df21655aca0b73497e5ea74c4c347c66085def05263136741aae4c476fdd
+size 128459314

pre_trained_model/whisper-ppg/D_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e7bfff64704b51c8f49d23fee8e292a47ac0b4dbf9887ebd5f867abf9353dc33
+size 187027205

pre_trained_model/whisper-ppg/G_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b5ff28db8fa5894fdcb29c8aff760e45c9fd3b88a892391dcf0d0257e80a78b1
+size 237719813

pretrain/checkpoint_best_legacy_500.pt

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

pretrain/fcpe.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c3a8dd2dbd51baf19ed295006f2ac25dba6dd60adc7ec578ae5fbd94970951da
+size 69005189

pretrain/hubert-soft-0d54a1f4.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e82e7d079df05fe3aa535f6f7d42d309bdae1d2a53324e2b2386c56721f4f649
+size 378435957

pretrain/medium.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:345ae4da62f9b3d59415adc60127b97c714f32e89e936602e85993674d08dcb1
+size 1528008539

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"]

pretrain/nsf_hifigan/NOTICE.txt

@@ -0,0 +1,74 @@
+--- DiffSinger Community Vocoder ---
+
+ARCHITECTURE: NSF-HiFiGAN
+RELEASE DATE: 2022-12-11
+
+HYPER PARAMETERS:
+ - 44100 sample rate
+ - 128 mel bins
+ - 512 hop size
+ - 2048 window size
+ - fmin at 40Hz
+ - fmax at 16000Hz
+
+
+NOTICE:
+
+All model weights in the [DiffSinger Community Vocoder Project](https://openvpi.github.io/vocoders/), including
+model weights in this directory, are provided by the [OpenVPI Team](https://github.com/openvpi/), under the
+[Attribution-NonCommercial-ShareAlike 4.0 International](https://creativecommons.org/licenses/by-nc-sa/4.0/) license.
+
+
+ACKNOWLEDGEMENTS:
+
+Training data of this vocoder is provided and permitted by the following organizations, societies and individuals:
+
+孙飒              https://www.qfssr.cn
+赤松_Akamatsu     https://www.zhibin.club
+乐威              https://www.zhibin.club
+伯添              https://space.bilibili.com/24087011
+雲宇光             https://space.bilibili.com/660675050
+橙子言             https://space.bilibili.com/318486464
+人衣大人           https://space.bilibili.com/2270344
+玖蝶              https://space.bilibili.com/676771003
+Yuuko
+白夜零BYL          https://space.bilibili.com/1605040503
+嗷天              https://space.bilibili.com/5675252
+洛泠羽            https://space.bilibili.com/347373318
+灰条纹的灰猫君      https://space.bilibili.com/2083633
+幽寂              https://space.bilibili.com/478860
+恶魔王女           https://space.bilibili.com/2475098
+AlexYHX 芮晴
+绮萱              https://y.qq.com/n/ryqq/singer/003HjD6H4aZn1K
+诗芸              https://y.qq.com/n/ryqq/singer/0005NInj142zm0
+汐蕾              https://y.qq.com/n/ryqq/singer/0023cWMH1Bq1PJ
+1262917464
+炜阳
+叶卡yolka
+幸の夏            https://space.bilibili.com/1017297686
+暮色未量           https://space.bilibili.com/272904686
+晓寞sama          https://space.bilibili.com/3463394
+没头绪的节操君
+串串BunC          https://space.bilibili.com/95817834
+落雨              https://space.bilibili.com/1292427
+长尾巴的翎艾        https://space.bilibili.com/1638666
+声闻计划           https://space.bilibili.com/392812269
+唐家大小姐         http://5sing.kugou.com/palmusic/default.html
+不伊子
+
+Training machines are provided by:
+
+花儿不哭           https://space.bilibili.com/5760446
+
+
+TERMS OF REDISTRIBUTIONS:
+
+1. Do not sell this vocoder, or charge any fees from redistributing it, as prohibited by
+   the license.
+2. Include a copy of the CC BY-NC-SA 4.0 license, or a link referring to it.
+3. Include a copy of this notice, or any other notices informing that this vocoder is
+   provided by the OpenVPI Team, that this vocoder is licensed under CC BY-NC-SA 4.0, and
+   with a complete acknowledgement list as shown above.
+4. If you fine-tuned or modified the weights, leave a notice about what has been changed.
+5. (Optional) Leave a link to the official release page of the vocoder, and tell users
+   that other versions and future updates of this vocoder can be obtained from the website.

pretrain/nsf_hifigan/NOTICE.zh-CN.txt

@@ -0,0 +1,72 @@
+--- DiffSinger 社区声码器 ---
+
+架构：NSF-HiFiGAN
+发布日期：2022-12-11
+
+超参数：
+ - 44100 sample rate
+ - 128 mel bins
+ - 512 hop size
+ - 2048 window size
+ - fmin at 40Hz
+ - fmax at 16000Hz
+
+
+注意事项：
+
+[DiffSinger 社区声码器企划](https://openvpi.github.io/vocoders/) 中的所有模型权重，
+包括此目录下的模型权重，均由 [OpenVPI Team](https://github.com/openvpi/) 提供，并基于
+[Attribution-NonCommercial-ShareAlike 4.0 International](https://creativecommons.org/licenses/by-nc-sa/4.0/)
+进行许可。
+
+
+致谢：
+
+此声码器的训练数据由以下组织、社团和个人提供并许可：
+
+孙飒              https://www.qfssr.cn
+赤松_Akamatsu     https://www.zhibin.club
+乐威              https://www.zhibin.club
+伯添              https://space.bilibili.com/24087011
+雲宇光             https://space.bilibili.com/660675050
+橙子言             https://space.bilibili.com/318486464
+人衣大人           https://space.bilibili.com/2270344
+玖蝶              https://space.bilibili.com/676771003
+Yuuko
+白夜零BYL          https://space.bilibili.com/1605040503
+嗷天              https://space.bilibili.com/5675252
+洛泠羽            https://space.bilibili.com/347373318
+灰条纹的灰猫君      https://space.bilibili.com/2083633
+幽寂              https://space.bilibili.com/478860
+恶魔王女           https://space.bilibili.com/2475098
+AlexYHX 芮晴
+绮萱              https://y.qq.com/n/ryqq/singer/003HjD6H4aZn1K
+诗芸              https://y.qq.com/n/ryqq/singer/0005NInj142zm0
+汐蕾              https://y.qq.com/n/ryqq/singer/0023cWMH1Bq1PJ
+1262917464
+炜阳
+叶卡yolka
+幸の夏            https://space.bilibili.com/1017297686
+暮色未量           https://space.bilibili.com/272904686
+晓寞sama          https://space.bilibili.com/3463394
+没头绪的节操君
+串串BunC          https://space.bilibili.com/95817834
+落雨              https://space.bilibili.com/1292427
+长尾巴的翎艾        https://space.bilibili.com/1638666
+声闻计划           https://space.bilibili.com/392812269
+唐家大小姐         http://5sing.kugou.com/palmusic/default.html
+不伊子
+
+训练算力的提供者如下：
+
+花儿不哭           https://space.bilibili.com/5760446
+
+
+二次分发条款：
+
+1. 请勿售卖此声码器或从其二次分发过程中收取任何费用，因为此类行为受到许可证的禁止。
+2. 请在二次分发文件中包含一份 CC BY-NC-SA 4.0 许可证的副本或指向该许可证的链接。
+3. 请在二次分发文件中包含这份声明，或以其他形式声明此声码器由 OpenVPI Team 提供并基于 CC BY-NC-SA 4.0 许可，
+   并附带上述完整的致谢名单。
+4. 如果您微调或修改了权重，请留下一份关于其受到了何种修改的说明。
+5.（可选）留下一份指向此声码器的官方发布页面的链接，并告知使用者可从该网站获取此声码器的其他版本和未来的更新。

pretrain/nsf_hifigan/config.json

@@ -0,0 +1,38 @@
+{
+    "resblock": "1",
+    "num_gpus": 4,
+    "batch_size": 10,
+    "learning_rate": 0.0002,
+    "adam_b1": 0.8,
+    "adam_b2": 0.99,
+    "lr_decay": 0.999,
+    "seed": 1234,
+
+    "upsample_rates":        [ 8, 8, 2, 2, 2],
+    "upsample_kernel_sizes": [16,16, 4, 4, 4],
+    "upsample_initial_channel": 512,
+    "resblock_kernel_sizes": [3,7,11],
+    "resblock_dilation_sizes": [[1,3,5], [1,3,5], [1,3,5]],
+    "discriminator_periods": [3, 5, 7, 11, 17, 23, 37],
+
+    "segment_size": 16384,
+    "num_mels": 128,
+    "num_freq": 1025,
+    "n_fft"   : 2048,
+    "hop_size": 512,
+    "win_size": 2048,
+
+    "sampling_rate": 44100,
+
+    "fmin": 40,
+    "fmax": 16000,
+    "fmax_for_loss": null,
+
+    "num_workers": 16,
+
+    "dist_config": {
+        "dist_backend": "nccl",
+        "dist_url": "tcp://localhost:54321",
+        "world_size": 1
+    }
+}

pretrain/nsf_hifigan/model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2c576b63b7ed952161b70fad34e0562ace502ce689195520d8a2a6c051de29d6
+size 56825430

pretrain/rmvpe.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d49bd662038808878c9d7420e0f583f506fe69086cc384f0da88f0b3a4e1115
+size 368492925