Compare commits
No commits in common. "09b6315ee9e0d51f9aea535958cba8456c54f3b5" and "d00782515bc1a53f1f3fc661ec9a7554c0336ed8" have entirely different histories.
09b6315ee9
...
d00782515b
|
|
@ -44,7 +44,7 @@
|
|||
|
||||
https://github.com/user-attachments/assets/10cbcc5f-4e70-45cf-bdc5-d6361e40bb86
|
||||
|
||||
[🔗 在线体验 (Gradio)](https://modelscope.cn/studios/gongjy/MiniMind-O) | [🔗 视频介绍](https://www.bilibili.com/video/BV1V1RsBcEMX)
|
||||
[🔗 在线体验](https://modelscope.cn/studios/gongjy/MiniMind-O) | [🔗 视频介绍](https://www.bilibili.com/video/BV1V1RsBcEMX)
|
||||
|
||||
|
||||
</div>
|
||||
|
|
@ -624,8 +624,6 @@ https://github.com/user-attachments/assets/3e8d0a76-282d-4a9d-9726-a954cf80198a
|
|||
- [SLAM-Omni](https://aclanthology.org/2025.findings-acl.115/)(数据)
|
||||
- [SenseVoice](https://arxiv.org/abs/2407.04051)(组件)
|
||||
- [Mimi / Moshi](https://arxiv.org/abs/2410.00037)(组件)
|
||||
- [vLLM-Omni](https://github.com/vllm-project/vllm-omni)(推理、合成数据)
|
||||
- 其他参考的开源项目与论文(在技术报告中详细列出)
|
||||
|
||||
## 🫶支持者
|
||||
|
||||
|
|
|
|||
|
|
@ -44,7 +44,7 @@
|
|||
|
||||
https://github.com/user-attachments/assets/10cbcc5f-4e70-45cf-bdc5-d6361e40bb86
|
||||
|
||||
[🔗 Online Demo (Gradio)](https://modelscope.cn/studios/gongjy/MiniMind-O) | [🔗 Video Intro](https://www.bilibili.com/video/BV1V1RsBcEMX)
|
||||
[🔗 Online Demo](https://modelscope.cn/studios/gongjy/MiniMind-O) | [🔗 Video Intro](https://www.bilibili.com/video/BV1V1RsBcEMX)
|
||||
|
||||
</div>
|
||||
|
||||
|
|
@ -622,8 +622,6 @@ That said, the value of MiniMind-O lies exactly here. It compresses an entire Om
|
|||
- [SLAM-Omni](https://aclanthology.org/2025.findings-acl.115/) (data)
|
||||
- [SenseVoice](https://arxiv.org/abs/2407.04051) (component)
|
||||
- [Mimi / Moshi](https://arxiv.org/abs/2410.00037) (component)
|
||||
- [vLLM-Omni](https://github.com/vllm-project/vllm-omni) (inference, synthetic data)
|
||||
- Other referenced open-source projects and papers (full list in the technical report)
|
||||
|
||||
## 🫶 Supporters
|
||||
|
||||
|
|
|
|||
|
|
@ -451,8 +451,7 @@ def init_model(args):
|
|||
load_main_model(M['models'][model_name], model_name)
|
||||
try:
|
||||
from transformers import MimiModel
|
||||
M['mimi'] = MimiModel.from_pretrained('../model/mimi').eval().to(args.device)
|
||||
if args.device != 'cpu': M['mimi'] = M['mimi'].half()
|
||||
M['mimi'] = MimiModel.from_pretrained('../model/mimi').eval()
|
||||
print('Mimi model loaded')
|
||||
except: M['mimi'] = None
|
||||
try:
|
||||
|
|
@ -492,12 +491,12 @@ def init_model(args):
|
|||
|
||||
if __name__ == '__main__':
|
||||
p = argparse.ArgumentParser()
|
||||
p.add_argument('--load_from', default='./', help='模型权重搜索目录;目录下可放多个 HF 格式模型,WebUI 会自动扫描并允许切换。')
|
||||
p.add_argument('--device', default='cuda' if torch.cuda.is_available() else 'cpu', help='推理设备;CUDA 可用时默认 cuda。显存不足或排查环境问题时可改为 cpu。')
|
||||
p.add_argument('--port', default=7860, type=int, help='WebUI 服务端口;端口被占用或需要同时启动多个实例时调整。')
|
||||
p.add_argument('--audio_chunk_frames', default=4, type=int, help='流式播放每次解码的 Mimi frame 数;默认 4 约 320ms。WebUI 播放卡顿时可调大到 8/12,低延迟优先时保持 4。')
|
||||
p.add_argument('--audio_overlap', default=2, type=int, help='分块 Mimi 解码的重叠帧数;默认 2 用于缓解块边界断裂。一般不需要调整,边界杂音明显时可适当增大。')
|
||||
p.add_argument('--max_history_turns', default=0, type=int, help='对话历史轮数;默认 0 不带历史以降低延迟和显存。需要多轮上下文时调大,但会增加 prefill 成本。')
|
||||
p.add_argument('--load_from', default='./')
|
||||
p.add_argument('--device', default='cuda' if torch.cuda.is_available() else 'cpu')
|
||||
p.add_argument('--port', default=7860, type=int)
|
||||
p.add_argument('--audio_chunk_frames', default=4, type=int)
|
||||
p.add_argument('--audio_overlap', default=2, type=int)
|
||||
p.add_argument('--max_history_turns', default=0, type=int)
|
||||
args = p.parse_args()
|
||||
init_model(args)
|
||||
app.run(host='0.0.0.0', port=args.port, threaded=True)
|
||||
|
|
|
|||
Loading…
Reference in New Issue