Add BERT + JVNV speaker models (PyTorch & ONNX)

.gitattributes

@@ -33,3 +33,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+_bert_cache/models--ku-nlp--deberta-v2-large-japanese-char-wwm/blobs/2630f547d018524a7b03506a42c700cbac49e29bdc441845b0615bfb3b5d74d2 filter=lfs diff=lfs merge=lfs -text
+onnx/bert/fp32/bert.onnx.data filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,3 +1,221 @@
----
-license: agpl-3.0
----
+---
+license: agpl-3.0
+language:
+  - ja
+tags:
+  - tts
+  - text-to-speech
+  - japanese
+  - onnx
+  - style-bert-vits2
+library_name: hayakoe
+---
+
+# HayaKoe Models
+
+## KR | 한국어
+
+HayaKoe의 사전학습 모델 저장소입니다. Style-Bert-VITS2 기반 일본어 TTS 모델을 PyTorch와 ONNX 형식으로 제공합니다.
+
+### 구조
+
+```
+pytorch/
+├── bert/           # DeBERTa v2 (ku-nlp/deberta-v2-large-japanese-char-wwm)
+│   ├── fp32/       # 원본 (1.3GB)
+│   ├── fp16/       # 반정밀도 (629MB) — VRAM 절감용
+│   └── q8/         # INT8 동적 양자화 (413MB) — CPU 메모리 절감용
+└── speakers/       # JVNV 화자 모델 (safetensors)
+    ├── jvnv-F1-jp/
+    ├── jvnv-F2-jp/
+    ├── jvnv-M1-jp/
+    └── jvnv-M2-jp/
+
+onnx/
+├── bert/           # ONNX 변환 BERT
+│   ├── fp32/       # FP32 ONNX (1.2GB)
+│   └── q8/         # INT8 양자화 (498MB)
+└── speakers/       # ONNX 변환 Synthesizer
+    └── jvnv-*/
+        ├── synthesizer.onnx      # FP32 (239MB)
+        └── synthesizer_q8.onnx   # INT8 (239MB)
+```
+
+### 추천 조합
+
+| 용도 | BERT | Synthesizer | 특징 |
+|------|------|-------------|------|
+| **빠른 CPU 추론** | `onnx/bert/q8` | `onnx/speakers/*/synthesizer.onnx` | 속도 1.6x, RAM 81% 절감 |
+| **GPU 추론** | `pytorch/bert/fp16` | `pytorch/speakers/*` | VRAM 40% 절감 |
+| **연구/실험** | `pytorch/bert/fp32` | `pytorch/speakers/*` | 원본 그대로 |
+
+### Synthesizer FP16 미제공 안내
+
+Synthesizer(VITS)의 Flow 레이어에 포함된 `rational_quadratic_spline` 함수가 반정밀도(FP16)에서 discriminant < 0 assertion 오류를 발생시킵니다. 따라서 **Synthesizer는 반드시 FP32로 사용**해야 하며, FP16 변환 모델은 제공하지 않습니다.
+
+### 화자 (JVNV)
+
+[JVNV Corpus](https://sites.google.com/site/shinaborumonosagashi/jvnv-corpus) 기반 사전학습 화자 모델입니다. 파인튜닝 없이 바로 추론에 사용할 수 있습니다.
+
+| 화자 | 성별 | 스타일 |
+|------|------|--------|
+| jvnv-F1-jp | 여성 | Neutral, Angry, Disgust, Fear, Happy, Sad, Surprise |
+| jvnv-F2-jp | 여성 | 동일 |
+| jvnv-M1-jp | 남성 | 동일 |
+| jvnv-M2-jp | 남성 | 동일 |
+
+---
+
+## EN | English
+
+Pre-trained model repository for HayaKoe, a fast ONNX-optimized Japanese TTS based on Style-Bert-VITS2.
+
+### Structure
+
+```
+pytorch/
+├── bert/           # DeBERTa v2 (ku-nlp/deberta-v2-large-japanese-char-wwm)
+│   ├── fp32/       # Original (1.3GB)
+│   ├── fp16/       # Half precision (629MB) — reduces VRAM
+│   └── q8/         # INT8 dynamic quantization (413MB) — reduces CPU RAM
+└── speakers/       # JVNV speaker models (safetensors)
+
+onnx/
+├── bert/           # ONNX-converted BERT
+│   ├── fp32/       # FP32 ONNX (1.2GB)
+│   └── q8/         # INT8 quantized (498MB)
+└── speakers/       # ONNX-converted Synthesizer
+    └── jvnv-*/
+        ├── synthesizer.onnx      # FP32 (239MB)
+        └── synthesizer_q8.onnx   # INT8 (239MB)
+```
+
+### Recommended Configurations
+
+| Use Case | BERT | Synthesizer | Notes |
+|----------|------|-------------|-------|
+| **Fast CPU inference** | `onnx/bert/q8` | `onnx/speakers/*/synthesizer.onnx` | 1.6x speedup, 81% RAM reduction |
+| **GPU inference** | `pytorch/bert/fp16` | `pytorch/speakers/*` | 40% VRAM reduction |
+| **Research** | `pytorch/bert/fp32` | `pytorch/speakers/*` | Original weights |
+
+### Why No Synthesizer FP16?
+
+The `rational_quadratic_spline` function in the VITS Flow layer triggers a discriminant < 0 assertion error under half precision (FP16). **Synthesizer must remain FP32.** No FP16 variant is provided.
+
+### Speakers (JVNV)
+
+Pre-trained speaker models from the [JVNV Corpus](https://sites.google.com/site/shinaborumonosagashi/jvnv-corpus). Ready for inference without fine-tuning.
+
+| Speaker | Gender | Styles |
+|---------|--------|--------|
+| jvnv-F1-jp | Female | Neutral, Angry, Disgust, Fear, Happy, Sad, Surprise |
+| jvnv-F2-jp | Female | Same |
+| jvnv-M1-jp | Male | Same |
+| jvnv-M2-jp | Male | Same |
+
+---
+
+## JP | 日本語
+
+HayaKoeの事前学習モデルリポジトリです。Style-Bert-VITS2ベースの日本語TTSモデルをPyTorchとONNX形式で提供します。
+
+### 構成
+
+```
+pytorch/
+├── bert/           # DeBERTa v2 (ku-nlp/deberta-v2-large-japanese-char-wwm)
+│   ├── fp32/       # オリジナル (1.3GB)
+│   ├── fp16/       # 半精度 (629MB) — VRAM削減用
+│   └── q8/         # INT8動的量子化 (413MB) — CPUメモリ削減用
+└── speakers/       # JVNV話者モデル (safetensors)
+
+onnx/
+├── bert/           # ONNX変換BERT
+│   ├── fp32/       # FP32 ONNX (1.2GB)
+│   └── q8/         # INT8量子化 (498MB)
+└── speakers/       # ONNX変換Synthesizer
+    └─��� jvnv-*/
+        ├── synthesizer.onnx      # FP32 (239MB)
+        └── synthesizer_q8.onnx   # INT8 (239MB)
+```
+
+### 推奨構成
+
+| 用途 | BERT | Synthesizer | 特徴 |
+|------|------|-------------|------|
+| **高速CPU推論** | `onnx/bert/q8` | `onnx/speakers/*/synthesizer.onnx` | 速度1.6倍、RAM 81%削減 |
+| **GPU推論** | `pytorch/bert/fp16` | `pytorch/speakers/*` | VRAM 40%削減 |
+| **研究** | `pytorch/bert/fp32` | `pytorch/speakers/*` | オリジナル |
+
+### Synthesizer FP16非対応について
+
+VITSのFlowレイヤーに含まれる`rational_quadratic_spline`関数が半精度(FP16)で判別式 < 0のアサーションエラーを発生させます。**Synthesizerは必ずFP32で使用**してください。FP16変換モデルは提供しません。
+
+### 話者 (JVNV)
+
+[JVNVコーパス](https://sites.google.com/site/shinaborumonosagashi/jvnv-corpus)ベースの事前学習話者モデルです。ファインチューニングなしで推論に使用できます。
+
+| 話者 | 性別 | スタイル |
+|------|------|----------|
+| jvnv-F1-jp | 女性 | Neutral, Angry, Disgust, Fear, Happy, Sad, Surprise |
+| jvnv-F2-jp | 女性 | 同上 |
+| jvnv-M1-jp | 男性 | 同上 |
+| jvnv-M2-jp | 男性 | 同上 |
+
+---
+
+## ZH | 中文
+
+HayaKoe的预训练模型仓库。提供基于Style-Bert-VITS2的日语TTS模型，支持PyTorch和ONNX格式。
+
+### 结构
+
+```
+pytorch/
+├── bert/           # DeBERTa v2 (ku-nlp/deberta-v2-large-japanese-char-wwm)
+│   ├── fp32/       # 原始 (1.3GB)
+│   ├── fp16/       # 半精度 (629MB) — 减少VRAM
+│   └── q8/         # INT8动态量化 (413MB) — 减少CPU内存
+└── speakers/       # JVNV说话人模型 (safetensors)
+
+onnx/
+├── bert/           # ONNX转换BERT
+│   ├── fp32/       # FP32 ONNX (1.2GB)
+│   └── q8/         # INT8量化 (498MB)
+└── speakers/       # ONNX转换Synthesizer
+    └── jvnv-*/
+        ├── synthesizer.onnx      # FP32 (239MB)
+        └── synthesizer_q8.onnx   # INT8 (239MB)
+```
+
+### 推荐配置
+
+| 用途 | BERT | Synthesizer | 特点 |
+|------|------|-------------|------|
+| **快速CPU推理** | `onnx/bert/q8` | `onnx/speakers/*/synthesizer.onnx` | 速度1.6倍，RAM减少81% |
+| **GPU推理** | `pytorch/bert/fp16` | `pytorch/speakers/*` | VRAM减少40% |
+| **研究** | `pytorch/bert/fp32` | `pytorch/speakers/*` | 原始权重 |
+
+### 为什么没有Synthesizer FP16？
+
+VITS Flow层中的`rational_quadratic_spline`函数在半精度(FP16)下会触发判别式 < 0的断言错误。**Synthesizer必须使用FP32**，不提供FP16转换模型。
+
+### 说话人 (JVNV)
+
+基于[JVNV语料库](https://sites.google.com/site/shinaborumonosagashi/jvnv-corpus)的预训练说话人模型，无需微调即可直接用于推理。
+
+| 说话人 | 性别 | 风格 |
+|--------|------|------|
+| jvnv-F1-jp | 女 | Neutral, Angry, Disgust, Fear, Happy, Sad, Surprise |
+| jvnv-F2-jp | 女 | 同上 |
+| jvnv-M1-jp | 男 | 同上 |
+| jvnv-M2-jp | 男 | 同上 |
+
+---
+
+## License
+
+- **HayaKoe**: AGPL-3.0
+- **BERT (DeBERTa v2)**: CC-BY-SA-4.0 ([ku-nlp/deberta-v2-large-japanese-char-wwm](https://huggingface.co/ku-nlp/deberta-v2-large-japanese-char-wwm))
+- **JVNV Speaker Models**: CC-BY-SA-4.0 ([litagin/style_bert_vits2_jvnv](https://huggingface.co/litagin/style_bert_vits2_jvnv))
+- **Style-Bert-VITS2**: AGPL-3.0 ([litagin02/Style-Bert-VITS2](https://github.com/litagin02/Style-Bert-VITS2))

_bert_cache/CACHEDIR.TAG

@@ -0,0 +1,4 @@
+Signature: 8a477f597d28d172789f06886806bc55
+# This file is a cache directory tag created by huggingface_hub.
+# For information about cache directory tags, see:
+#	https://bford.info/cachedir/

_bert_cache/models--ku-nlp--deberta-v2-large-japanese-char-wwm/blobs/2630f547d018524a7b03506a42c700cbac49e29bdc441845b0615bfb3b5d74d2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2630f547d018524a7b03506a42c700cbac49e29bdc441845b0615bfb3b5d74d2
+size 1318453704

_bert_cache/models--ku-nlp--deberta-v2-large-japanese-char-wwm/blobs/8ab2175580e45760875557201e5543019ca3039b

@@ -0,0 +1,19 @@
+{
+  "cls_token": "[CLS]",
+  "do_lower_case": false,
+  "do_subword_tokenize": true,
+  "do_word_tokenize": true,
+  "jumanpp_kwargs": null,
+  "mask_token": "[MASK]",
+  "mecab_kwargs": null,
+  "model_max_length": 1000000000000000019884624838656,
+  "never_split": null,
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "special_tokens_map_file": null,
+  "subword_tokenizer_type": "character",
+  "sudachi_kwargs": null,
+  "tokenizer_class": "BertJapaneseTokenizer",
+  "unk_token": "[UNK]",
+  "word_tokenizer_type": "basic"
+}

_bert_cache/models--ku-nlp--deberta-v2-large-japanese-char-wwm/blobs/9fb6b0ac2ec49b6556e58b5ed9492eb33166714d

@@ -0,0 +1,37 @@
+{
+  "architectures": [
+    "DebertaV2ForMaskedLM"
+  ],
+  "attention_head_size": 64,
+  "attention_probs_dropout_prob": 0.1,
+  "conv_act": "gelu",
+  "conv_kernel_size": 3,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 1024,
+  "initializer_range": 0.02,
+  "intermediate_size": 4096,
+  "layer_norm_eps": 1e-07,
+  "max_position_embeddings": 512,
+  "max_relative_positions": -1,
+  "model_type": "deberta-v2",
+  "norm_rel_ebd": "layer_norm",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 24,
+  "pad_token_id": 0,
+  "pooler_dropout": 0,
+  "pooler_hidden_act": "gelu",
+  "pooler_hidden_size": 1024,
+  "pos_att_type": [
+    "p2c",
+    "c2p"
+  ],
+  "position_biased_input": false,
+  "position_buckets": 256,
+  "relative_attention": true,
+  "share_att_key": true,
+  "torch_dtype": "float16",
+  "transformers_version": "4.25.1",
+  "type_vocab_size": 0,
+  "vocab_size": 22012
+}

_bert_cache/models--ku-nlp--deberta-v2-large-japanese-char-wwm/blobs/a8b3208c2884c4efb86e49300fdd3dc877220cdf

@@ -0,0 +1,7 @@
+{
+  "cls_token": "[CLS]",
+  "mask_token": "[MASK]",
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "unk_token": "[UNK]"
+}

_bert_cache/models--ku-nlp--deberta-v2-large-japanese-char-wwm/refs/main

@@ -0,0 +1 @@
+547b0e8b044fba3f9b84d0ab9f990440bd130c8b

_bert_cache/models--ku-nlp--deberta-v2-large-japanese-char-wwm/snapshots/547b0e8b044fba3f9b84d0ab9f990440bd130c8b/config.json

@@ -0,0 +1,37 @@
+{
+  "architectures": [
+    "DebertaV2ForMaskedLM"
+  ],
+  "attention_head_size": 64,
+  "attention_probs_dropout_prob": 0.1,
+  "conv_act": "gelu",
+  "conv_kernel_size": 3,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 1024,
+  "initializer_range": 0.02,
+  "intermediate_size": 4096,
+  "layer_norm_eps": 1e-07,
+  "max_position_embeddings": 512,
+  "max_relative_positions": -1,
+  "model_type": "deberta-v2",
+  "norm_rel_ebd": "layer_norm",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 24,
+  "pad_token_id": 0,
+  "pooler_dropout": 0,
+  "pooler_hidden_act": "gelu",
+  "pooler_hidden_size": 1024,
+  "pos_att_type": [
+    "p2c",
+    "c2p"
+  ],
+  "position_biased_input": false,
+  "position_buckets": 256,
+  "relative_attention": true,
+  "share_att_key": true,
+  "torch_dtype": "float16",
+  "transformers_version": "4.25.1",
+  "type_vocab_size": 0,
+  "vocab_size": 22012
+}

_bert_cache/models--ku-nlp--deberta-v2-large-japanese-char-wwm/snapshots/547b0e8b044fba3f9b84d0ab9f990440bd130c8b/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2630f547d018524a7b03506a42c700cbac49e29bdc441845b0615bfb3b5d74d2
+size 1318453704

_bert_cache/models--ku-nlp--deberta-v2-large-japanese-char-wwm/snapshots/547b0e8b044fba3f9b84d0ab9f990440bd130c8b/special_tokens_map.json

@@ -0,0 +1,7 @@
+{
+  "cls_token": "[CLS]",
+  "mask_token": "[MASK]",
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "unk_token": "[UNK]"
+}

_bert_cache/models--ku-nlp--deberta-v2-large-japanese-char-wwm/snapshots/547b0e8b044fba3f9b84d0ab9f990440bd130c8b/tokenizer_config.json

@@ -0,0 +1,19 @@
+{
+  "cls_token": "[CLS]",
+  "do_lower_case": false,
+  "do_subword_tokenize": true,
+  "do_word_tokenize": true,
+  "jumanpp_kwargs": null,
+  "mask_token": "[MASK]",
+  "mecab_kwargs": null,
+  "model_max_length": 1000000000000000019884624838656,
+  "never_split": null,
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "special_tokens_map_file": null,
+  "subword_tokenizer_type": "character",
+  "sudachi_kwargs": null,
+  "tokenizer_class": "BertJapaneseTokenizer",
+  "unk_token": "[UNK]",
+  "word_tokenizer_type": "basic"
+}

onnx/bert/fp32/bert.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:65670b756e4fd410b2b52b224a826cd401dce3f8579ae6b64c55febd95c272be
+size 267696

onnx/bert/fp32/bert.onnx.data

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:953e3136e020bba48fb84fcf55894d44ad96eb3198fc1b6087588c15ce9d5cfd
+size 1213382656

onnx/bert/fp32/config.json

@@ -0,0 +1,41 @@
+{
+  "architectures": [
+    "DebertaV2ForMaskedLM"
+  ],
+  "attention_head_size": 64,
+  "attention_probs_dropout_prob": 0.1,
+  "bos_token_id": null,
+  "conv_act": "gelu",
+  "conv_kernel_size": 3,
+  "dtype": "float32",
+  "eos_token_id": null,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 1024,
+  "initializer_range": 0.02,
+  "intermediate_size": 4096,
+  "layer_norm_eps": 1e-07,
+  "legacy": true,
+  "max_position_embeddings": 512,
+  "max_relative_positions": -1,
+  "model_type": "deberta-v2",
+  "norm_rel_ebd": "layer_norm",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 24,
+  "pad_token_id": 0,
+  "pooler_dropout": 0,
+  "pooler_hidden_act": "gelu",
+  "pooler_hidden_size": 1024,
+  "pos_att_type": [
+    "p2c",
+    "c2p"
+  ],
+  "position_biased_input": false,
+  "position_buckets": 256,
+  "relative_attention": true,
+  "share_att_key": true,
+  "tie_word_embeddings": true,
+  "transformers_version": "5.5.0",
+  "type_vocab_size": 0,
+  "vocab_size": 22012
+}

onnx/bert/fp32/tokenizer_config.json

@@ -0,0 +1,62 @@
+{
+  "added_tokens_decoder": {
+    "0": {
+      "content": "[PAD]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "[CLS]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "[SEP]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "3": {
+      "content": "[UNK]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "4": {
+      "content": "[MASK]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "backend": "custom",
+  "cls_token": "[CLS]",
+  "do_lower_case": false,
+  "do_subword_tokenize": true,
+  "do_word_tokenize": true,
+  "is_local": false,
+  "jumanpp_kwargs": null,
+  "mask_token": "[MASK]",
+  "mecab_kwargs": null,
+  "model_max_length": 1000000000000000019884624838656,
+  "never_split": null,
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "subword_tokenizer_type": "character",
+  "sudachi_kwargs": null,
+  "tokenizer_class": "BertJapaneseTokenizer",
+  "unk_token": "[UNK]",
+  "word_tokenizer_type": "basic"
+}

onnx/bert/q8/bert_q8.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1c0e17e0d30a8d1d371638055cace156c80ed2af98c0afdb9cef6057e3161cf7
+size 521580327

onnx/bert/q8/config.json

@@ -0,0 +1,41 @@
+{
+  "architectures": [
+    "DebertaV2ForMaskedLM"
+  ],
+  "attention_head_size": 64,
+  "attention_probs_dropout_prob": 0.1,
+  "bos_token_id": null,
+  "conv_act": "gelu",
+  "conv_kernel_size": 3,
+  "dtype": "float32",
+  "eos_token_id": null,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 1024,
+  "initializer_range": 0.02,
+  "intermediate_size": 4096,
+  "layer_norm_eps": 1e-07,
+  "legacy": true,
+  "max_position_embeddings": 512,
+  "max_relative_positions": -1,
+  "model_type": "deberta-v2",
+  "norm_rel_ebd": "layer_norm",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 24,
+  "pad_token_id": 0,
+  "pooler_dropout": 0,
+  "pooler_hidden_act": "gelu",
+  "pooler_hidden_size": 1024,
+  "pos_att_type": [
+    "p2c",
+    "c2p"
+  ],
+  "position_biased_input": false,
+  "position_buckets": 256,
+  "relative_attention": true,
+  "share_att_key": true,
+  "tie_word_embeddings": true,
+  "transformers_version": "5.5.0",
+  "type_vocab_size": 0,
+  "vocab_size": 22012
+}

onnx/bert/q8/tokenizer_config.json

@@ -0,0 +1,62 @@
+{
+  "added_tokens_decoder": {
+    "0": {
+      "content": "[PAD]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "[CLS]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "[SEP]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "3": {
+      "content": "[UNK]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "4": {
+      "content": "[MASK]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "backend": "custom",
+  "cls_token": "[CLS]",
+  "do_lower_case": false,
+  "do_subword_tokenize": true,
+  "do_word_tokenize": true,
+  "is_local": false,
+  "jumanpp_kwargs": null,
+  "mask_token": "[MASK]",
+  "mecab_kwargs": null,
+  "model_max_length": 1000000000000000019884624838656,
+  "never_split": null,
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "subword_tokenizer_type": "character",
+  "sudachi_kwargs": null,
+  "tokenizer_class": "BertJapaneseTokenizer",
+  "unk_token": "[UNK]",
+  "word_tokenizer_type": "basic"
+}

onnx/speakers/jvnv-F1-jp/config.json

@@ -0,0 +1,92 @@
+{
+  "train": {
+    "log_interval": 200,
+    "eval_interval": 1000,
+    "seed": 42,
+    "epochs": 300,
+    "learning_rate": 0.0001,
+    "betas": [0.8, 0.99],
+    "eps": 1e-9,
+    "batch_size": 4,
+    "bf16_run": true,
+    "fp16_run": false,
+    "lr_decay": 0.99996,
+    "segment_size": 16384,
+    "init_lr_ratio": 1,
+    "warmup_epochs": 0,
+    "c_mel": 45,
+    "c_kl": 1.0,
+    "c_commit": 100,
+    "skip_optimizer": true,
+    "freeze_ZH_bert": false,
+    "freeze_JP_bert": false,
+    "freeze_EN_bert": false,
+    "freeze_emo": false,
+    "freeze_style": false
+  },
+  "data": {
+    "use_jp_extra": true,
+    "training_files": "Data/jvnv-F1-jp/train.list",
+    "validation_files": "Data/jvnv-F1-jp/val.list",
+    "max_wav_value": 32768.0,
+    "sampling_rate": 44100,
+    "filter_length": 2048,
+    "hop_length": 512,
+    "win_length": 2048,
+    "n_mel_channels": 128,
+    "mel_fmin": 0.0,
+    "mel_fmax": null,
+    "add_blank": true,
+    "n_speakers": 1,
+    "cleaned_text": true,
+    "spk2id": {
+      "jvnv-F1-jp": 0
+    },
+    "num_styles": 7,
+    "style2id": {
+      "Neutral": 0,
+      "Angry": 1,
+      "Disgust": 2,
+      "Fear": 3,
+      "Happy": 4,
+      "Sad": 5,
+      "Surprise": 6
+    }
+  },
+  "model": {
+    "use_spk_conditioned_encoder": true,
+    "use_noise_scaled_mas": true,
+    "use_mel_posterior_encoder": false,
+    "use_duration_discriminator": false,
+    "use_wavlm_discriminator": true,
+    "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, 8, 2, 2],
+    "n_layers_q": 3,
+    "use_spectral_norm": false,
+    "gin_channels": 512,
+    "slm": {
+      "model": "./slm/wavlm-base-plus",
+      "sr": 16000,
+      "hidden": 768,
+      "nlayers": 13,
+      "initial_channel": 64
+    }
+  },
+  "version": "2.0-JP-Extra",
+  "model_name": "jvnv-F1-jp"
+}

onnx/speakers/jvnv-F1-jp/style_vectors.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1f959bb45ed0922efc31ff24e9147253814f42cb1d2d1e2bb10391a9df368489
+size 7296

onnx/speakers/jvnv-F1-jp/synthesizer.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5fb5013a0b7e57c3a50f2bf53046feb2fab1f34de454bdb81f386567eb367171
+size 250644697

onnx/speakers/jvnv-F1-jp/synthesizer_q8.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d9faffebfc22333eebba25dea271204c2843d06f6371dba624b51a4f7ec6ad6a
+size 250072350

onnx/speakers/jvnv-F2-jp/config.json

@@ -0,0 +1,92 @@
+{
+  "train": {
+    "log_interval": 200,
+    "eval_interval": 1000,
+    "seed": 42,
+    "epochs": 300,
+    "learning_rate": 0.0001,
+    "betas": [0.8, 0.99],
+    "eps": 1e-9,
+    "batch_size": 4,
+    "bf16_run": false,
+    "fp16_run": false,
+    "lr_decay": 0.99996,
+    "segment_size": 16384,
+    "init_lr_ratio": 1,
+    "warmup_epochs": 0,
+    "c_mel": 45,
+    "c_kl": 1.0,
+    "c_commit": 100,
+    "skip_optimizer": true,
+    "freeze_ZH_bert": false,
+    "freeze_JP_bert": false,
+    "freeze_EN_bert": false,
+    "freeze_emo": false,
+    "freeze_style": false
+  },
+  "data": {
+    "use_jp_extra": true,
+    "training_files": "/content/drive/MyDrive/Style-Bert-VITS2/Data/jvnv-F2/train.list",
+    "validation_files": "/content/drive/MyDrive/Style-Bert-VITS2/Data/jvnv-F2/val.list",
+    "max_wav_value": 32768.0,
+    "sampling_rate": 44100,
+    "filter_length": 2048,
+    "hop_length": 512,
+    "win_length": 2048,
+    "n_mel_channels": 128,
+    "mel_fmin": 0.0,
+    "mel_fmax": null,
+    "add_blank": true,
+    "n_speakers": 1,
+    "cleaned_text": true,
+    "spk2id": {
+      "jvnv-F2-jp": 0
+    },
+    "num_styles": 7,
+    "style2id": {
+      "Neutral": 0,
+      "Angry": 1,
+      "Disgust": 2,
+      "Fear": 3,
+      "Happy": 4,
+      "Sad": 5,
+      "Surprise": 6
+    }
+  },
+  "model": {
+    "use_spk_conditioned_encoder": true,
+    "use_noise_scaled_mas": true,
+    "use_mel_posterior_encoder": false,
+    "use_duration_discriminator": false,
+    "use_wavlm_discriminator": true,
+    "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, 8, 2, 2],
+    "n_layers_q": 3,
+    "use_spectral_norm": false,
+    "gin_channels": 512,
+    "slm": {
+      "model": "./slm/wavlm-base-plus",
+      "sr": 16000,
+      "hidden": 768,
+      "nlayers": 13,
+      "initial_channel": 64
+    }
+  },
+  "version": "2.0-JP-Extra",
+  "model_name": "jvnv-F2-jp"
+}

onnx/speakers/jvnv-F2-jp/style_vectors.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:900f8cde3a336d12193fec7b7d8e6c5dc77b3a5d719a9be3f8598389cd88e643
+size 7296

onnx/speakers/jvnv-F2-jp/synthesizer.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b3679affc29b6de52c2dd006c032e8900ba731b73f990f9ac7b5b4ec59634afe
+size 250644697

onnx/speakers/jvnv-F2-jp/synthesizer_q8.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:480db251beb0818cd20aac02c6007ea9cdffafa007dcc75777452e4596b4ddf5
+size 250072350

onnx/speakers/jvnv-M1-jp/config.json

@@ -0,0 +1,92 @@
+{
+  "train": {
+    "log_interval": 200,
+    "eval_interval": 1000,
+    "seed": 42,
+    "epochs": 300,
+    "learning_rate": 0.0001,
+    "betas": [0.8, 0.99],
+    "eps": 1e-9,
+    "batch_size": 4,
+    "bf16_run": true,
+    "fp16_run": false,
+    "lr_decay": 0.99996,
+    "segment_size": 16384,
+    "init_lr_ratio": 1,
+    "warmup_epochs": 0,
+    "c_mel": 45,
+    "c_kl": 1.0,
+    "c_commit": 100,
+    "skip_optimizer": true,
+    "freeze_ZH_bert": false,
+    "freeze_JP_bert": false,
+    "freeze_EN_bert": false,
+    "freeze_emo": false,
+    "freeze_style": false
+  },
+  "data": {
+    "use_jp_extra": true,
+    "training_files": "Data/jvnv-M1-jp/train.list",
+    "validation_files": "Data/jvnv-M1-jp/val.list",
+    "max_wav_value": 32768.0,
+    "sampling_rate": 44100,
+    "filter_length": 2048,
+    "hop_length": 512,
+    "win_length": 2048,
+    "n_mel_channels": 128,
+    "mel_fmin": 0.0,
+    "mel_fmax": null,
+    "add_blank": true,
+    "n_speakers": 1,
+    "cleaned_text": true,
+    "spk2id": {
+      "jvnv-M1-jp": 0
+    },
+    "num_styles": 7,
+    "style2id": {
+      "Neutral": 0,
+      "Angry": 1,
+      "Disgust": 2,
+      "Fear": 3,
+      "Happy": 4,
+      "Sad": 5,
+      "Surprise": 6
+    }
+  },
+  "model": {
+    "use_spk_conditioned_encoder": true,
+    "use_noise_scaled_mas": true,
+    "use_mel_posterior_encoder": false,
+    "use_duration_discriminator": false,
+    "use_wavlm_discriminator": true,
+    "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, 8, 2, 2],
+    "n_layers_q": 3,
+    "use_spectral_norm": false,
+    "gin_channels": 512,
+    "slm": {
+      "model": "./slm/wavlm-base-plus",
+      "sr": 16000,
+      "hidden": 768,
+      "nlayers": 13,
+      "initial_channel": 64
+    }
+  },
+  "version": "2.0-JP-Extra",
+  "model_name": "jvnv-M1-jp"
+}

onnx/speakers/jvnv-M1-jp/style_vectors.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7a925435e8c1c9efc8fc8e90e690655ab9a7bae00a790892e13e936510d04f05
+size 7296

onnx/speakers/jvnv-M1-jp/synthesizer.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:33833654608a95ebe8322ab48ba761b9e2ffe53f336f8df4381efd5fe383ca58
+size 250644697

onnx/speakers/jvnv-M1-jp/synthesizer_q8.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1ff46f5eff7a5e550eae8f76bf27e3b4aaf1c043a8b9879ff5a14753a44cca89
+size 250072350

onnx/speakers/jvnv-M2-jp/config.json

@@ -0,0 +1,92 @@
+{
+  "train": {
+    "log_interval": 200,
+    "eval_interval": 1000,
+    "seed": 42,
+    "epochs": 300,
+    "learning_rate": 0.0001,
+    "betas": [0.8, 0.99],
+    "eps": 1e-9,
+    "batch_size": 4,
+    "bf16_run": true,
+    "fp16_run": false,
+    "lr_decay": 0.99996,
+    "segment_size": 16384,
+    "init_lr_ratio": 1,
+    "warmup_epochs": 0,
+    "c_mel": 45,
+    "c_kl": 1.0,
+    "c_commit": 100,
+    "skip_optimizer": true,
+    "freeze_ZH_bert": false,
+    "freeze_JP_bert": false,
+    "freeze_EN_bert": false,
+    "freeze_emo": false,
+    "freeze_style": false
+  },
+  "data": {
+    "use_jp_extra": true,
+    "training_files": "Data/jvnv-M2-jp/train.list",
+    "validation_files": "Data/jvnv-M2-jp/val.list",
+    "max_wav_value": 32768.0,
+    "sampling_rate": 44100,
+    "filter_length": 2048,
+    "hop_length": 512,
+    "win_length": 2048,
+    "n_mel_channels": 128,
+    "mel_fmin": 0.0,
+    "mel_fmax": null,
+    "add_blank": true,
+    "n_speakers": 1,
+    "cleaned_text": true,
+    "spk2id": {
+      "jvnv-M2-jp": 0
+    },
+    "num_styles": 7,
+    "style2id": {
+      "Neutral": 0,
+      "Angry": 1,
+      "Disgust": 2,
+      "Fear": 3,
+      "Happy": 4,
+      "Sad": 5,
+      "Surprise": 6
+    }
+  },
+  "model": {
+    "use_spk_conditioned_encoder": true,
+    "use_noise_scaled_mas": true,
+    "use_mel_posterior_encoder": false,
+    "use_duration_discriminator": false,
+    "use_wavlm_discriminator": true,
+    "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, 8, 2, 2],
+    "n_layers_q": 3,
+    "use_spectral_norm": false,
+    "gin_channels": 512,
+    "slm": {
+      "model": "./slm/wavlm-base-plus",
+      "sr": 16000,
+      "hidden": 768,
+      "nlayers": 13,
+      "initial_channel": 64
+    }
+  },
+  "version": "2.0-JP-Extra",
+  "model_name": "jvnv-M2-jp"
+}

onnx/speakers/jvnv-M2-jp/style_vectors.npy

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c965bb63fa4a759d41a8a4a3649333125d6497ae8a705d81b7d5c5bd2854797c
+size 7296

onnx/speakers/jvnv-M2-jp/synthesizer.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e82e01385c9f8240c520a0134f17df870298357aa70340e135440d70fe701f1
+size 250644697

onnx/speakers/jvnv-M2-jp/synthesizer_q8.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cbbf9082a7933a521cfa02b3e19c44441bd24f17cf37c68d4f50ba6b8436bf7c
+size 250072350

pytorch/bert/fp16/config.json

@@ -0,0 +1,41 @@
+{
+  "architectures": [
+    "DebertaV2ForMaskedLM"
+  ],
+  "attention_head_size": 64,
+  "attention_probs_dropout_prob": 0.1,
+  "bos_token_id": null,
+  "conv_act": "gelu",
+  "conv_kernel_size": 3,
+  "dtype": "float16",
+  "eos_token_id": null,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 1024,
+  "initializer_range": 0.02,
+  "intermediate_size": 4096,
+  "layer_norm_eps": 1e-07,
+  "legacy": true,
+  "max_position_embeddings": 512,
+  "max_relative_positions": -1,
+  "model_type": "deberta-v2",
+  "norm_rel_ebd": "layer_norm",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 24,
+  "pad_token_id": 0,
+  "pooler_dropout": 0,
+  "pooler_hidden_act": "gelu",
+  "pooler_hidden_size": 1024,
+  "pos_att_type": [
+    "p2c",
+    "c2p"
+  ],
+  "position_biased_input": false,
+  "position_buckets": 256,
+  "relative_attention": true,
+  "share_att_key": true,
+  "tie_word_embeddings": true,
+  "transformers_version": "5.5.0",
+  "type_vocab_size": 0,
+  "vocab_size": 22012
+}

pytorch/bert/fp16/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a40217f1e566c480300036a0f630524cec161bf2e69d053cf39f2ec08dab7fd3
+size 659250232