Spaces:
Running
on
T4
How to enable Seamless expressive?
I got this warning in the console:
[WARNING][simuleval_transcoder]: Passing 'expressive' but the agent does not support expressive output!
I want to replicate the same scenario as this space but locally, let me know if you have any ideas.
Thanks!
@Encinashael
This means that you need to download the expressive model, please follow instructions here to fill out the request form: https://github.com/facebookresearch/seamless_communication/?tab=readme-ov-file#seamlessexpressive-models. Then you can put the model in models/Seamless/pretssel_melhifigan_wm.pt, see https://huggingface.co/spaces/facebook/seamless-streaming/commit/d9b3f79412fac72fc173419d769f7979c47674cf for more details
@Encinashael This means that you need to download the expressive model, please follow instructions here to fill out the request form: https://github.com/facebookresearch/seamless_communication/?tab=readme-ov-file#seamlessexpressive-models. Then you can put the model in
models/Seamless/pretssel_melhifigan_wm.pt, see https://huggingface.co/spaces/facebook/seamless-streaming/commit/d9b3f79412fac72fc173419d769f7979c47674cf for more details
Hi! I have completed the form and downloaded the models. I've put them on those directories but still get the same error "[WARNING][simuleval_transcoder]: Passing 'expressive' but the agent does not support expressive output!" when checking "expressive" in the web
The downloaded files names where:
m2m_expressive_unity.pt
pretssel_melhifigan_wm-16khz.pt
pretssel_melhifigan_wm-final.pt
I also changed pretssel_melhifigan_wm-final.pt to pretssel_melhifigan_wm.pt but still doesn't work.
I have them both in the /seamless-streaming/seamless_server/models/Seamless/ and /seamless-streaming/seamless_server/models/SeamlessStreaming/ directories but it doens't work :(
Hi!
The default model is SeamlessStreaming.
So just copy the .yaml file from the Seamless model to the SeamlessStreaming folder
I ran into the same problem, it was an issue with the expressive models not being in the correct place as @annasun28 pointed out here
The environment config is being set in the run_docker.sh
If you trace and debug the conditionals where the warning is being sent it should get you on the right path.