1. 首页
  2. 问答
  3. [index-tts]如何提前下载好所有模型，需要纯离线不联网运行

[index-tts]如何提前下载好所有模型，需要纯离线不联网运行

2025-10-27 981 views
0
Image

图上的方案，是用国内hf镜像站代替， 但是我这需要的是不联网离线运行，能不能提前把需要的模型都下载好？

SimZhou

回答

5

肯定可以的，我是macos，先去把模型全下好了，然后放入文件夹中搞定了

TonyD0g

6

模型已经提前下载到本地，但是第二次启动报错： (MaxRetryError("HTTPSConnectionPool(host='hf-mirror.com', port=443): Max retries exceeded with url: /facebook/w2v-bert-2.0/resolve/main/preprocessor_config.json (Caused by SSLError(SSLCertVerificationError(1, '[SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: unable to get local issuer certificate (_ssl.c:1007)')))"), '(Request ID: ec92f6aa-2d1d-439c-9c4b-a4c4b6d407b1)')' thrown while requesting HEAD https://hf-mirror.com/facebook/w2v-bert-2.0/resolve/main/preprocessor_config.json 官方说法是还会下载一些小模型。 为什么不能一步到位？

Redroadsl

1

模型已经提前下载到本地，但是第二次启动报错： (MaxRetryError("HTTPSConnectionPool(host='hf-mirror.com', port=443): Max retries exceeded with url: /facebook/w2v-bert-2.0/resolve/main/preprocessor_config.json (Caused by SSLError(SSLCertVerificationError(1, '[SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: unable to get local issuer certificate (_ssl.c:1007)')))"), '(Request ID: ec92f6aa-2d1d-439c-9c4b-a4c4b6d407b1)')' thrown while requesting HEAD https://hf-mirror.com/facebook/w2v-bert-2.0/resolve/main/preprocessor_config.json 官方说法是还会下载一些小模型。 为什么不能一步到位？

这就不知道了，我反正是看缺哪个模型就自己去下载好放入指定文件夹中

TonyD0g

9

我这边已经成功了。分享给各位：

需要下载的模型有：

# ==== download IndexTTS-2 =====
modelscope download --model IndexTeam/IndexTTS-2             --local_dir /home/yihua.zhou/models/IndexTeam/IndexTTS-2
sed -ri 's# "nvidia/bigvgan_v2_22khz_80band_256x"# "/home/yihua.zhou/models/nvidia/bigvgan_v2_22khz_80band_256x"#g' /home/yihua.zhou/models/IndexTeam/IndexTTS-2/config.yaml
# modelscope download --model IndexTeam/IndexTTS-1.5         --local_dir /home/yihua.zhou/models/IndexTeam/IndexTTS-1.5
# modelscope download --model IndexTeam/Index-TTS            --local_dir /home/yihua.zhou/models/IndexTeam/Index-TTS

# ==== download https://huggingface.co/facebook/w2v-bert-2.0 into models/facebook/w2v-bert-v2.0 =====
hf download facebook/w2v-bert-2.0                            --local-dir /home/yihua.zhou/models/facebook/w2v-bert-2.0
hf download amphion/MaskGCT semantic_codec/model.safetensors --local-dir /home/yihua.zhou/models/amphion/MaskGCT
hf download funasr/campplus                                  --local-dir /home/yihua.zhou/models/funasr/campplus
hf download nvidia/bigvgan_v2_22khz_80band_256x bigvgan_generator.pt config.json --local-dir /home/yihua.zhou/models/nvidia/bigvgan_v2_22khz_80band_256x
hf download Plachta/JDCnet bst.t7                            --local-dir /home/yihua.zhou/models/Plachta/JDCnet

然后，需要手动修改代码中写死的模型路径默认值， 以下这些文件都需要改：

Image

具体我就不把每个文件打开看了，大家进去看一下，from_pretained()函数里写的路径，还有什么 os.path()，很多都是仓库名，改成你下载的绝对路径就好。随便举一个例子：

Image

SimZhou

4

我这边已经成功了。分享给各位：

需要下载的模型有：

==== download IndexTTS-2 =====

modelscope download --model IndexTeam/IndexTTS-2 --local_dir /home/yihua.zhou/models/IndexTeam/IndexTTS-2 sed -ri 's# "nvidia/bigvgan_v2_22khz_80band_256x"# "/home/yihua.zhou/models/nvidia/bigvgan_v2_22khz_80band_256x"#g' /home/yihua.zhou/models/IndexTeam/IndexTTS-2/config.yaml

modelscope download --model IndexTeam/IndexTTS-1.5 --local_dir /home/yihua.zhou/models/IndexTeam/IndexTTS-1.5 modelscope download --model IndexTeam/Index-TTS --local_dir /home/yihua.zhou/models/IndexTeam/Index-TTS ==== download https://huggingface.co/facebook/w2v-bert-2.0 into models/facebook/w2v-bert-v2.0 =====

hf download facebook/w2v-bert-2.0 --local-dir /home/yihua.zhou/models/facebook/w2v-bert-2.0 hf download amphion/MaskGCT semantic_codec/model.safetensors --local-dir /home/yihua.zhou/models/amphion/MaskGCT hf download funasr/campplus --local-dir /home/yihua.zhou/models/funasr/campplus hf download nvidia/bigvgan_v2_22khz_80band_256x bigvgan_generator.pt config.json --local-dir /home/yihua.zhou/models/nvidia/bigvgan_v2_22khz_80band_256x hf download Plachta/JDCnet bst.t7 --local-dir /home/yihua.zhou/models/Plachta/JDCnet 然后，需要手动修改代码中写死的模型路径默认值， 以下这些文件都需要改： Image

具体我就不把每个文件打开看了，大家进去看一下，from_pretained()函数里写的路径，还有什么 os.path()，很多都是仓库名，改成你下载的绝对路径就好。随便举一个例子：

Image

可否将修改的内容导出为git的diff文件，方便我们知道改哪些地方。

ganecheng

3

@SimZhou

ganecheng

6

我这边已经成功了。分享给各位： 需要下载的模型有：

==== download IndexTTS-2 =====

modelscope download --model IndexTeam/IndexTTS-2 --local_dir /home/yihua.zhou/models/IndexTeam/IndexTTS-2 sed -ri 's# "nvidia/bigvgan_v2_22khz_80band_256x"# "/home/yihua.zhou/models/nvidia/bigvgan_v2_22khz_80band_256x"#g' /home/yihua.zhou/models/IndexTeam/IndexTTS-2/config.yaml

modelscope download --model IndexTeam/IndexTTS-1.5 --local_dir /home/yihua.zhou/models/IndexTeam/IndexTTS-1.5 modelscope download --model IndexTeam/Index-TTS --local_dir /home/yihua.zhou/models/IndexTeam/Index-TTS ==== download https://huggingface.co/facebook/w2v-bert-2.0 into models/facebook/w2v-bert-v2.0 =====

hf download facebook/w2v-bert-2.0 --local-dir /home/yihua.zhou/models/facebook/w2v-bert-2.0 hf download amphion/MaskGCT semantic_codec/model.safetensors --local-dir /home/yihua.zhou/models/amphion/MaskGCT hf download funasr/campplus --local-dir /home/yihua.zhou/models/funasr/campplus hf download nvidia/bigvgan_v2_22khz_80band_256x bigvgan_generator.pt config.json --local-dir /home/yihua.zhou/models/nvidia/bigvgan_v2_22khz_80band_256x hf download Plachta/JDCnet bst.t7 --local-dir /home/yihua.zhou/models/Plachta/JDCnet 然后，需要手动修改代码中写死的模型路径默认值， 以下这些文件都需要改： Image 具体我就不把每个文件打开看了，大家进去看一下，from_pretained()函数里写的路径，还有什么 os.path()，很多都是仓库名，改成你下载的绝对路径就好。随便举一个例子：

Image

可否将修改的内容导出为git的diff文件，方便我们知道改哪些地方。

请笑纳： diff.txt

SimZhou

7

实测，直接执行以下命令提前下载模型就行，不用修改代码中模型的路径

uv tool install "huggingface-hub[cli,hf_xet]"

hf download IndexTeam/IndexTTS-2 --local-dir=checkpoints

hf download facebook/w2v-bert-2.0
hf download amphion/MaskGCT semantic_codec/model.safetensors
hf download funasr/campplus
hf download nvidia/bigvgan_v2_22khz_80band_256x bigvgan_generator.pt config.json
hf download Plachta/JDCnet bst.t7

可构建出完全离线使用的Dockerfile可参考 index-tts-2.zip

ganecheng