docs: update requirements.txt #217

- 主要是修复 pydantic 更新问题，并增加所有主要依赖的版本限制

Author: zhzLuke96

README.en.md

@@ -137,8 +137,6 @@ Since Forge primarily focuses on API functionality development, automatic downlo
 
 > **Note**: If you need to use ModelScope to download models, use `--source modelscope`. Some models may not be available for download using ModelScope.
 
-> **About CosyVoice**: It's unclear which model to use. Overall, the `instruct` model has the most features, but its quality may not be the best. If you wish to use other models, feel free to select `dl_cosyvoice_base.py`, `dl_cosyvoice_instruct.py`, or the `sft` script. The loading priority is `base` > `instruct` > `sft`, and you can determine which to load based on folder existence.
-
 ## FAQ
 
 ### How to perform voice cloning?

README.md

@@ -17,7 +17,7 @@ Speech-AI-Forge 是一个围绕 TTS 生成模型开发的项目，实现了 API
 
 ## Installation and Running
 
-首先，确保 [相关依赖](./docs/dependencies.md) 已经正确安装
+首先，确保 [相关依赖](./docs/dependencies.md) 已经正确安装，并查看 [模型下载](#模型下载) 下载所需模型
 
 启动：
 
@@ -141,8 +141,6 @@ WIP 开发中
 
 > **注意**：如果需要使用 ModelScope 下载模型，请使用 `--source modelscope`。部分模型可能无法使用 ModelScope 下载。
 
-> **关于 CosyVoice**：不太确定应该使用哪个模型。整体来看，`instruct` 模型功能最多，但可能质量不是最佳。如果需要使用其他模型，请自行选择 `dl_cosyvoice_base.py`、`dl_cosyvoice_instruct.py` 或 `sft` 脚本。加载优先级为 `base` > `instruct` > `sft`，可根据文件夹存在性判断加载顺序。
-
 ## FAQ
 
 ### 如何语音复刻？

docs/dependencies.md

@@ -146,7 +146,7 @@ apt-get install rubberband-cli
 > （如果直接运行某些情况可能会安装 cpu 版本，具体应该指定什么版本请自行确定）
 
 ```bash
-pip3 install torch torchvision torchaudio --index-url https://mirror.sjtu.edu.cn/pytorch-wheels/cu121
+pip3 install torch==2.3.1 torchvision torchaudio --index-url https://mirror.sjtu.edu.cn/pytorch-wheels/cu121
 ```
 
 如果上面无法安装 cuda 版本，可以参考下面的指令安装指定版本 torch (linux)
@@ -167,14 +167,6 @@ pip3 install https://mirror.sjtu.edu.cn/pytorch-wheels/cu121/torchvision-0.18.1%
 python -m pip install -r requirements.txt
 ```
 
-对于 windows 环境，请使用 `requirements.windows.txt`
-
-> 因为有类似 `pynini` 这样的可选依赖库不支持在 windows 环境安装
-
-```bash
-python -m pip install -r requirements.windows.txt
-```
-
 ## 6. 其他
 
 如果需要部署 Docker 环境，请配置 Docker 和 Docker Compose。

requirements.txt

@@ -1,88 +1,98 @@
 numpy==1.26.4
-scipy
-lxml
-pydub
-fastapi
-soundfile
-omegaconf
-pypinyin
-vocos
-pandas
-vector_quantize_pytorch
-einops
-transformers~=4.41.1
+scipy==1.11.4
+lxml==5.2.2
+pydub==0.25.1
+fastapi==0.111.0
+soundfile==0.12.1
+omegaconf==2.3.0
+pypinyin==0.51.0
+vocos==0.1.0
+pandas==2.1.3
+vector_quantize_pytorch==1.14.24
+einops==0.8.0
+transformers==4.41.2
 huggingface-hub==0.23.2
-omegaconf~=2.3.0
-tqdm
-python-ms
-# torch
-# torchvision
-# torchaudio
+tqdm==4.66.3
+python-ms==1.1.0
 
 # colab
-datasets
+datasets==2.20.0
 
 # TN
-emojiswitch
-python-dotenv
-zhon
+emojiswitch==0.0.3
+python-dotenv==1.0.1
+zhon==2.0.2
 mistune==3.0.2
 
 # gradio
 gradio==4.44.0
 gradio_client==1.3.0
+## NOTE: pydantic 不要升到 2.11 会报错！
+pydantic==2.8.2 
+pydantic-core==2.20.1
 
 # audio_denoiser
-python-box
-ftfy
-librosa
-pyrubberband
-pybase16384
-cachetools
-langdetect
-WeTextProcessing
+python-box==7.1.1
+ftfy==6.2.0
+librosa==0.10.1
+pyrubberband==0.3.0
+pybase16384==0.3.8
+cachetools==5.3.3
+langdetect==1.0.9
+WeTextProcessing; sys_platform=="linux"
 
 #### fish-speech
-hydra-core
-loguru
-lightning
-natsort
-pyrootutils
-loralib
-wandb
-
+hydra-core==1.3.2
+loguru==0.7.2
+lightning==2.3.3
+natsort==8.4.0
+pyrootutils==1.0.4
+loralib==0.1.2
+wandb==0.17.4
+# NOTE: 安装这个是为了保证没有 Import error 其实是用不到的，所以版本可能不对
+accelerate==1.6.0
 
 ### cosyvoice
-hyperpyyaml
-openai-whisper
-onnxruntime
-inflect
-matcha-tts
+hyperpyyaml==1.2.2
+openai-whisper==20231117
+onnxruntime==1.18.1
+inflect==7.3.1
+unidecode==1.3.8
+# NOTE: 这个似乎不需要
+# matcha-tts
 
 # whisper
 faster_whisper==1.0.3
 
 # webui
-pysubs2
+pysubs2==1.7.3
 
 # OpenVoice
-eng-to-ipa
-jieba
-cn2an
+eng-to-ipa==0.0.2
+jieba==0.42.1
+cn2an==0.5.22
 
 ### fire red tts
-lingua-language-detector
+lingua-language-detector==2.0.2
+diffusers==0.32.2
 
 ### F5 TTS
-torchdiffeq
-x-transformers
-ema-pytorch
-funasr
-jiwer
-zhconv
+torchdiffeq==0.2.4
+x-transformers==1.39.1
+ema-pytorch==0.7.0
+funasr==1.1.12
+jiwer==3.0.4
+zhconv==1.4.3
+tomli==2.2.1
 
 ### webui tools
 moviepy==1.0.3
 
 ### downloader script
-modelscope
+modelscope==1.20.0
+
+### PyTorch Dependencies (using custom source for CUDA)
+--extra-index-url https://download.pytorch.org/whl/cu121
+torch==2.3.1+cu121
+torchvision==0.18.1+cu121
+torchaudio==2.3.1+cu121