Spaces:
Running
Running
What GPU are you using for the Gradio demo?
#3
by
Arete7
- opened
In the Gradio demo, it takes around two minutes to synthesize speech for this brief text: "Hello there, how are you doing today?"
This is super slow. What GPU are you using for the Gradio demo? And what is the recommended GPU for production use cases?
This space is running on CPU tier (2 vCPU, 16GB RAM) not GPU, so longer wait times should be expected. For recommended GPUs and solid performance, something in the NVIDIA 30 series should work well, especially if you use llama.cpp for generation.