Realtime-Venus local preparation commited on
Commit
ddbeee6
·
1 Parent(s): a1c592b

Refine public model documentation

Browse files
Files changed (2) hide show
  1. README.md +1 -1
  2. README_zh.md +2 -2
README.md CHANGED
@@ -705,4 +705,4 @@ libraries, and any data used with this checkpoint.
705
 
706
  Model release: `03b5c2919dd684900bdf6852272ac6a8f30fab74`. Root configuration: `15d5b420cd8557e9582d41a57a80af3e4de18e47`.
707
 
708
- Root `config.yaml` stores the complete Audio and Omni configurations separately. Each child `config.json` preserves AutoClass entry points and pins the root YAML commit and SHA-256 through `venus_config_ref`. Native AutoConfig loading expands this reference; weights, Tokenizer, Processor and TTS resources remain relative to the selected model directory. A mismatched current YAML falls back to the pinned version. Missing offline configuration and digest mismatches fail explicitly. Normal Hub caching is retained; full offline exports require the expanded config and all companion resources.
 
705
 
706
  Model release: `03b5c2919dd684900bdf6852272ac6a8f30fab74`. Root configuration: `15d5b420cd8557e9582d41a57a80af3e4de18e47`.
707
 
708
+ Root `config.yaml` contains the Audio and Omni configurations. Each model directory's `config.json` references a fixed version through `venus_config_ref`. Load the selected model directory with Transformers; weights, Tokenizer, Processor and TTS resources reside in that directory. For offline use, retain the configuration, custom code and all model resources.
README_zh.md CHANGED
@@ -127,7 +127,7 @@ hf download inclusionAI/Realtime-Venus Realtime-Venus-Omni/requirements.txt --re
127
  python -m pip install -r Realtime-Venus-Omni/requirements.txt
128
  ```
129
 
130
- 也可以在包含 `download_models.py` 的目录中,通过 Python 完成一次下载准备。它与上述命令使用同一个下载入口:
131
 
132
  ```python
133
  from download_models import download_models
@@ -616,4 +616,4 @@ print("Saved generated speech to output/audio_full_duplex.wav")
616
 
617
  模型发布版本:`03b5c2919dd684900bdf6852272ac6a8f30fab74`。根配置版本:`15d5b420cd8557e9582d41a57a80af3e4de18e47`。
618
 
619
- 根 `config.yaml` 分别保存 Audio/Omni 的完整配置。子目录 `config.json` 保留 AutoClass 入口,并通过 `venus_config_ref` 固定根 YAML 的提交号和 SHA-256。配置类在原生 AutoConfig 加载阶段展开引用,权重、Tokenizer、Processor、TTS 仍从所选模型子目录读取。当前快照的 YAML 不匹配时查找指定旧版本;离线缺少配置或摘要错误会明确失败。正常使用 Hub 缓存;完整导出保留展开后的配置及全部配套资源。
 
127
  python -m pip install -r Realtime-Venus-Omni/requirements.txt
128
  ```
129
 
130
+ 也可以在包含 `download_models.py` 的目录中,通过 Python 下载模型:
131
 
132
  ```python
133
  from download_models import download_models
 
616
 
617
  模型发布版本:`03b5c2919dd684900bdf6852272ac6a8f30fab74`。根配置版本:`15d5b420cd8557e9582d41a57a80af3e4de18e47`。
618
 
619
+ 根 `config.yaml` 保存 Audio/Omni 模型配置,子目录 `config.json` 通过 `venus_config_ref` 引用固定版本。使用 Transformers 从对应模型子目录加载,权重、Tokenizer、Processor 和 TTS 资源均位于该目录。离线使用需同时保留配置、自定义代码及模型配套资源。