Streaming implementation
#1
by
taresh18 - opened
Hi, I tested this model with sglang following the docs - https://github.com/sgl-project/sglang-omni/blob/main/sglang_omni/models/fishaudio_s2_pro/README.md
- I cant see any endpoint for streaming, is it not implemented?
- The emotion tags usage is pretty inconsistent, most of the times it just completely ignores the tags. Are there any guidelines on usage to get maximum accuracy?
- Is there a list of emotion tags which are stab;e and work better than others?
- I believe it's implemented, you may wanna create an issue in sglang-omni
2 / 3. TAG guideline will be available shortly